A few key points from an experience OEM torque control calibration engineer, if I may...
Throttle bodies are an example of compressible flow. It's not as simple as some people think. The best, complete definition can be found in Appendix C of
Heywood's textbook available here. In particular, figure C-3 shows the VERY nonlinear relationship between pressure ratio and mass flow rate below the choke flow point (0.528 PT/p0 as mentioned previously).
Somewhere inside the ECU, there is a table/model that correlates blade angle and
effective throttle area. If you have a different throttle body (ported, larger, whatever) then this should be adjusted. Unfortunately, we don't always have access to this table in our limited tuning software. (The Ford guys using HPT do though, go figure) This effective area bakes in the nonlinearity and makes it easier for the ECU to resolve a desired mass flow to a specific blade angle for that throttle body's geometry.
In GM vehicles, we have access to a single max throttle area scalar. Moving this number up for a larger throttle body works ONLY if the new throttle is an exact scaled up version of the stock geometry. If the aftermarket throttle manufacturer makes a larger bore, but plays games to make the low throttle positions "just like stock", this breaks the predicted relationship between low blade angles and effective flow area. A throttle relearn helps a little, but you still have a physical mismatch between ECU calibration and the hardware on the engine. You will always be chasing some error here and hoping the learns can drag it in line. Personally, I won't waste time trying to calibrate these unless I had access to both the ECU throttle area tables and complete flow bench data that resolves effective throttle area (like we do on the popular Ford throttle bodies).
By comparison, if one makes a larger throttle body that is simply scaled up, then all you need to do to calibrate it properly is adjust the max area scalar. No other learn is necessary because the ECU calibration now matches the hardware. I have done this with success before.
Any time you see large negative timing adjustments at idle, it's a symptom that the actual airflow is higher than desired. If the throttle body is leaving too much actual effective flow area open, actual airflow will be greater. It doesn't matter if you "tuned" the MAF or VVE to "fix" this, it's still happening. That "fix" would also have your idle fuel trims walking away from reality too, not great. Running very negative timing also erodes the range of available torque control for negative idle corrections (overspeed events), which becomes a circular problem. It also opens the door for poor combustion quality, incomplete burns, erratic O2 readings, and poor emissions.
From a calibrator's stand point, the proper solution is to have all aspects of the calibration follow the physics of the hardware. This means that both MAF and VVE should be independently accurate to actual airflow and the throttle body model must match the hardware on the engine without any nonlinear deviations. Anything else sets you up for chasing your tail and depending upon learned trims to adjust for calibration errors. A proper calibration and hardware match requires basically zero idle trims, just like when we get the fueling right and see only small STFT/LTFT adjustments.