Set KFKHFM to 0.66 from 140 load and KFLF to 1.5 from the same point.
Everything is fine now.

This is a pretty annoying limitation though. And I am sure it can be easily circumvented in ASM by doing some divisions by 2...

Wow. Thanks for actually tracking it down.

Someday we'll figure out how to either run an underscaled MAP, or fix the load limitation entirely

I didn't realize there was yet another limitation (ps_w) that wasn't based on MAP readings.

Would underscaling the MAP prevent ps_w from maxing out? Or is only calculated from load? I haven't checked the FR yet.

This works just fine. I don't see the point to under scale down low, as I would lose fine grained control and blow up part throttle.
Overscaling up top is fine, because this thing has a 1 bar wastegate spring and boost is 30 psi anyway, so the transition is really fast.
That said, I played with KFLF quite a bit just now on the airfield, and I got my fuel right where I want it, because I am linearizing the PID, I also had to do this with 0, 10, 20, 30, 40 etc duty cycle and actual follows requested nicely. This is a narrowband ECU as well.

As for fixing the problem entirely - it will need a simple ASM hack. Before rlroh_w goes into the integrator and gets transformed into ps_w and pbr_w, it should be divided by 2, then multiplied by 2 again.
ps_w is not important, what is important is pbr_w, and keeping it lower until we get back to rl_w. Underscaling the MAP will accomplish nothing at all, because ps_w is calculated from rlroh_w.

When I get some free time (usually in winter, when we have snow), then I will probably write this hack into the ECU. I rewrote tons of stuff on the older motronics to suit my needs as well - factory systems don't really like more than 200hp per liter

I didn't think about this solution! (
I agree keeping high resolution at low load is a good idea.
I may have to try it very soon.

BTW, which turbocharger are this logs referring to?

You can't fix this on a specific power output number. Other ECUs have some stepstones that limit you far before that
 
I have wraped my head about this limit as well because my new setup is also excceding the MAP limit and I don't like underscaled load because you fuck up nearly everything with that.
I think the most elegant solution would be to completely underscale the pressures as the flow over the throttle plate is a function of PR before - after the plate. So something similar like running at high altitude with a "bigger displacement" engine. To get it completely right you have to touch some stuff but it will run with no negative side effects.
When my hardware is done in 3weeks I'll start testing and let you know how to do that.

Thinking about this whole thing.

How many things would break just by setting KFURL 2x lower and adjusting the MAP sensor scaling 2x lower?
This would allow the ECU to read 4 bar boost, but at a double loss of precision.

Obviously the zero (atmospheric) point would have to be recalibrated and the relationship between the baro pressure and boost pressure, along with any axes that used boost/absolute pressure and maps, but I don't see anything impossible.

I will wait until my emulator arrives, and then try this on my RS4, and see how far I get before hitting a brick wall

do I find a short summary how the other 8-byte limits like MAP in ME7 have been solved?
(I will check with search function meanwhile).

thanks.