Restore your deleted fault classes and you will probably see the problem quite fast.

It's most likely caused by b_hfm being false and ECU going into alpha-n to msdk_w.
Because your msdk_w is calibrated poorly as well, you have a such jump.

In this case there is a correlation "pulling" between the two models.
And this is exactly caused by what you said. ps_w being much lower than pvdk_w at WOT.

You must calibrate your manifold model correctly, so that they are roughly equal at WOT.
Then correct msdk_w to follow mshfm using WDKUGDN.
You can of course disable this correlation/correction as well, but this is not really the correct way to go.

There are two such corrections. One from MSDK and one from PVDK, the limits are in these labels:
FKMSDKMN
FKMSDKMX
FKPVDKMN
FKPVDKMX

Look at BGMSZS module in FR.

For a quick fix just setting them all to 1 will solve your issue instantly.
But it's of course better to calibrate the model correctly.

Solved the issue yes, but you also defeated some protections in the process.
If your MAF is reading a little wrong during old age then this helps the load to be correct, so the car does not go too lean/run with too much timing or vice versa.

If you disable this correction, then the ECU trusts the MAF almost blind.