hmmm, the other time I saw some behavior like this was when Oldpoopie (Justin in Portland) found air bubbles in my fuel from a micro-cracked thermostatic T fitting. I can't remember if you'd addressed that already.
The QA (quantity adjuster) collar's feedback mechanism can't control for air in fuel, as metering is open-looped based on SOI and injection duration and assumed a 'normal' amount of dissolved gas. Bubbles really screw with fuel delivery.
Get a buddy to rev it while you look for bubbles. Wiggle the thermo-T (gently)...if you have bubbles, that can be a problem.
I also don't remember if you watched the VNT vacuum pot move the lever through full range while a buddy gets it on boost? They're known to get sticky, but that would show up on VCDS boost data of course.
What VCDS logging have you done so far?
You need to do fuel actual versus requested and boost actual versus requested on a 3rd gear pull.