Thanks to all replies. This is the expected result:
chrisb wrote: ↑
Sun Sep 13, 2020 9:44 pm
The culprit seems to be FinalDepth. If I set it to 0.001mm The profile switches sides.
This is a good hint! I only played with "Extra offset", which also effects the result, but not in a usable way. Actually I think "Final Depth" is calculated with "OpFinalDepth" which calculates to "-0.00". For me, it seems to be a rounding error somewhere. Using a clear "0" instead of "OpFinalDepth" solves the problem.
Originaly the solid is part of an assembly and therefore not located in z=0. It is moved to z=0 on the bottom side, but due to rounding (or maybe effective resolution) the bottom face is located at "-62nm" (nano!)
As nobody has declared this to a feature, I'm going to open a bug report.