Trying to use percentage corrections from TuneECU in a PCV, or vice-versa does not work. The numbers mean different things in each. TuneECU/TuneBoy use mg of air/ms, while PC units use ms of on-time. Injectors vary in how quickly they open, as well.
Ladies and gents, other than at cruise, the AFR tables in TuneECU or TuneBoy mean nothing, as they are out of range for the narrow-band sensor. Please do not use these in auto-tune tables.
Concerning auto-tune, you MUST realize a few limitations of Lambda sensors before you set up a table. First is that pump fuel DOES NOT have a stoich ratio of 14.7 as most think. Even NTK and Bosch sensors are calibrated for a stoich ratio of 14.56, and this is not correct. Most pump fuels in the USA contain 9-10% ethanol, and many of these fuels have a stoich ratio of 14.1-14.2 for Lambda 1.0
Next you need to know the actual lag time involved in different gears for the gasses to reach the sensor and the sensor to display accurately. This can mean that under hard acceleration, the numbers in the AFR trim table (used in auto-tune software) can lag behind actual engine speeds by 1500-2000 rpm in top gear, and much more in lower gears.
This means that the trim function can be trying to alter fueling at 5000 rpm based on samples at 3000 rpm. This is VERY IMPORTANT to get correct.
And finally, you need to understand that each injector may flow within 1-2 percent at 100% duty-cycle. The ECU can only handle ~86% d-c before the drivers in the ECU will shut down to prevent them from burning up. However, under 40% d-c, variation between injectors can easily be 14%, and I have seen as much as 42% differences in some rockets.
You need to give the engine WHAT IS WANTS, NOT WHAT YOU THINK IT WANTS. This is why "auto-tune" is BS.
I hope this helps,
-Wayne