Fuel trims negative after MAF and VE calibration. Actual lambda ratio also shows lean according to the wideband reading.
Tuned MAF and VE according to Greg Banish's methods. They come in consistently within a few percent when isolated.
If scanning and I turn off fuel trim learn, closed loop, and reset fuel trims via vcm controls my lambda error settles down to within a few percent. Assuming it is then fueling primarily off of the MAF/VE tables. Re-enable learn, and closed loop and the fuel trims quickly relearn a 7-15% error. The wideband reads lean at this point about that much as well. The car runs well either way. I have driven it around in forced open loop as well as trimming the 7-15%. Long terms in the histogram don't show more than 8% on average.
I have changed the General>fuel>air fuel ratio>AFR stoich to 14.124 (currently set at this) as well as down to 13.93 (10-15% ethanol fuel around here). The error remains the same. When toggled in VCM editor things improve. I can't seem to find a table or multiplier that could be causing this in VCM editor. I have thought about scaling my injectors, but don't think that is a proper way to fix the problem. What am I missing?
Current tune as well as a log file with marks toggling between fuel trim learn/closed loop off and learn on/open loop on via vcm controls.