Okay, so it has taken me probably two years to figure out how to tune DFCO in the 05-06 TJs without getting the stalling issue. Figured I would document my solution here so someone in the future can figure it out faster.
First off, what is Deceleration Fuel Cut-Off (DFCO)? Quite simply, it is an open-loop fueling condition in which the PCM completely shuts off the fuel injectors while the vehicle is decelerating. This saves fuel by reducing the amount of fuel the vehicle burns while decelerating in gear to nearly zero, and also makes the engine braking more aggressive. Without DFCO, you can expect fairly mild engine braking because the engine is still burning fuel/air, which provides some resistance to slowing down. When DFCO is engaged, 100% of the energy required to turn the engine is coming from the kinetic energy of the vehicle.
For the 05-06 TJs, it took me a very long time to figure out how to tune it.
So there are actually two different triggers for DFCO. One we can tune with HP Tuners, one we have no control over.
1st, DFCO is occasionally engaged in the TJ for emissions reasons. 3-way catalysts like the TJ has are designed to reduce carbon monoxide, hydrocarbons, and NOx emissions. CO is less relevant for this conversation, so I'll skip over it. As a broad oversimplification, your car tends to create more NOx emissions when running lean, and more hydrocarbon emissions when running rich. Hence why it normally tries to run stoichiometric. However, if it runs rich, such as when you enter PE or catalyst overtemp protection kicks in, it depletes the stored oxygen in the catalyst, and HC emissions begin to rise. The ECM corrects for this by trying to run DFCO the next time you let off the gas to recharge the catalytic converter with oxygen. (Complete fuel shut-off is used in lieu of simply leaning the mixture because it generates less NOx and also has less risk of detonation.) This implementation of DFCO we have no control over without outright disabling emissions altogether. However; it works well, so we just leave it be.
2nd, the PCMs are programmed to some degree to allow different fuel-saving strategies to be implemented, namely Multiple Displacement System (MDS) and DFCO. MDS I have not messed with; I suspect it needs certain hardware for it to work without destroying the catalytic converters. However, DFCO we do have limited control over. In the stock tune, it is disabled by setting the "Enable DFCO MAP" to an impossibly low manifold absolute pressure (MAP).
The root of the problem (I think) is that we don't have access to every parameter we need to be able to tune it from the factory. Either that, or it was so buggy even the Chrysler engineers gave up on it. What happens is that with settings similar to what you'd find on NGC4 or NGC4A-equipped vehicles, it works as expected 90% of the time. However, it seems in certain cases, the logic gets flipped, and it actually enables DFCO when it should be kicking out instead, and the vehicle stalls. Even more fun is that it doesn't restart normally. (More on this later.)
Factory settings:
Note that the Enable MAP is set to 2.6 kPa across the board. This effectively disables DFCO (except for emissions reasons) as MAP never goes below ~10 kPa in practice.
For comparison, here is a factory DFCO tune from a 2016 Jeep Wrangler using a GPEC2 ECM:
Note there are a lot more parameters here than in the NGC3. I suspect these actually are in the NGC3 programming; we just don't have access to them with HP Tuners.
After probably trying 50 different variations of all the variables in any way that made sense, I gave up on trying to use the DFCO settings as they were intended to be used. DFCO always worked great with every tune, but the stalling issue was just a constant thorn in my side. The biggest issue is that with the engine below the enable temperature, it would actually flip the logic. It wouldn't enable DFCO, but at the instant it would otherwise disable it, it would kick in and stall the engine.
Finally, on the 51st (plus or minus a dozen) attempt, I tried a final kamikaze tune and set the "Enable MAP" way above the "Disable MAP" and set it across the board to 60 kPa. Figured it would stall everything out immediately. Instead, it did the opposite. I couldn't get a single stall driving. But suddenly, the engine would die at idle if I let out the clutch too fast. And lowering that 60 kPa to 40 kPa made that far worse. So I doubled down on the Kamikaze tune:
And it works! Beautifully. And yes that enable ECT and that Maximum MAP settings are exactly where they need to be. Does it make sense looking at it? Heck no. But it works.
My best guess as to the error in the ECM logic is that the DFCO state is stored as a yes/no state, or effectively 1/0. If one parameter wants DFCO disabled, it flips and disables itself. But if two parameters (such as idle state, Disable MAP Hysteresis, Disable MAP or Enable ECT) then both say to disable the DFCO simultaneously, it flips it to enable DFCO, generally when you least want it.
So effectively we tune it so DFCO is always enabled, and only one parameter can kick it off at a time. If you set the Enable ECT too high, the logic will flip when the engine is cold, and you will stall. Set the maximum MAP too low, and the Disable MAP Hysteresis will line up with the idle request, and you will stall. But if DFCO is always enabled, it only ever gets kicked off, not re-enabled. So you don't ever stall.
First off, what is Deceleration Fuel Cut-Off (DFCO)? Quite simply, it is an open-loop fueling condition in which the PCM completely shuts off the fuel injectors while the vehicle is decelerating. This saves fuel by reducing the amount of fuel the vehicle burns while decelerating in gear to nearly zero, and also makes the engine braking more aggressive. Without DFCO, you can expect fairly mild engine braking because the engine is still burning fuel/air, which provides some resistance to slowing down. When DFCO is engaged, 100% of the energy required to turn the engine is coming from the kinetic energy of the vehicle.
For the 05-06 TJs, it took me a very long time to figure out how to tune it.
So there are actually two different triggers for DFCO. One we can tune with HP Tuners, one we have no control over.
1st, DFCO is occasionally engaged in the TJ for emissions reasons. 3-way catalysts like the TJ has are designed to reduce carbon monoxide, hydrocarbons, and NOx emissions. CO is less relevant for this conversation, so I'll skip over it. As a broad oversimplification, your car tends to create more NOx emissions when running lean, and more hydrocarbon emissions when running rich. Hence why it normally tries to run stoichiometric. However, if it runs rich, such as when you enter PE or catalyst overtemp protection kicks in, it depletes the stored oxygen in the catalyst, and HC emissions begin to rise. The ECM corrects for this by trying to run DFCO the next time you let off the gas to recharge the catalytic converter with oxygen. (Complete fuel shut-off is used in lieu of simply leaning the mixture because it generates less NOx and also has less risk of detonation.) This implementation of DFCO we have no control over without outright disabling emissions altogether. However; it works well, so we just leave it be.
2nd, the PCMs are programmed to some degree to allow different fuel-saving strategies to be implemented, namely Multiple Displacement System (MDS) and DFCO. MDS I have not messed with; I suspect it needs certain hardware for it to work without destroying the catalytic converters. However, DFCO we do have limited control over. In the stock tune, it is disabled by setting the "Enable DFCO MAP" to an impossibly low manifold absolute pressure (MAP).
The root of the problem (I think) is that we don't have access to every parameter we need to be able to tune it from the factory. Either that, or it was so buggy even the Chrysler engineers gave up on it. What happens is that with settings similar to what you'd find on NGC4 or NGC4A-equipped vehicles, it works as expected 90% of the time. However, it seems in certain cases, the logic gets flipped, and it actually enables DFCO when it should be kicking out instead, and the vehicle stalls. Even more fun is that it doesn't restart normally. (More on this later.)
Factory settings:
Note that the Enable MAP is set to 2.6 kPa across the board. This effectively disables DFCO (except for emissions reasons) as MAP never goes below ~10 kPa in practice.
For comparison, here is a factory DFCO tune from a 2016 Jeep Wrangler using a GPEC2 ECM:
Note there are a lot more parameters here than in the NGC3. I suspect these actually are in the NGC3 programming; we just don't have access to them with HP Tuners.
After probably trying 50 different variations of all the variables in any way that made sense, I gave up on trying to use the DFCO settings as they were intended to be used. DFCO always worked great with every tune, but the stalling issue was just a constant thorn in my side. The biggest issue is that with the engine below the enable temperature, it would actually flip the logic. It wouldn't enable DFCO, but at the instant it would otherwise disable it, it would kick in and stall the engine.
Finally, on the 51st (plus or minus a dozen) attempt, I tried a final kamikaze tune and set the "Enable MAP" way above the "Disable MAP" and set it across the board to 60 kPa. Figured it would stall everything out immediately. Instead, it did the opposite. I couldn't get a single stall driving. But suddenly, the engine would die at idle if I let out the clutch too fast. And lowering that 60 kPa to 40 kPa made that far worse. So I doubled down on the Kamikaze tune:
And it works! Beautifully. And yes that enable ECT and that Maximum MAP settings are exactly where they need to be. Does it make sense looking at it? Heck no. But it works.
My best guess as to the error in the ECM logic is that the DFCO state is stored as a yes/no state, or effectively 1/0. If one parameter wants DFCO disabled, it flips and disables itself. But if two parameters (such as idle state, Disable MAP Hysteresis, Disable MAP or Enable ECT) then both say to disable the DFCO simultaneously, it flips it to enable DFCO, generally when you least want it.
So effectively we tune it so DFCO is always enabled, and only one parameter can kick it off at a time. If you set the Enable ECT too high, the logic will flip when the engine is cold, and you will stall. Set the maximum MAP too low, and the Disable MAP Hysteresis will line up with the idle request, and you will stall. But if DFCO is always enabled, it only ever gets kicked off, not re-enabled. So you don't ever stall.
or 
or 
is many more facepalms than I've ever given, so Jeff is the winner! 
