Well, all I can further contribute is to say:
- Mesh leveling works for me (BLTouch v2, SKR 1.4T, TMC2209)
- If it would be a systematic thing, a lot more people would have complained
I appreciate your effort to tackle the problem. I’d recommend a more systematic approach:
- Use a bare minimum
klipper.cfg
to rule out any interference with other settings (just steppers, bltouch, mcu etc) - Verify probe repeatability under Marlin as well as under Klipper → both should show the same accuracy
- Compare Klipper’s mesh with Marlin’s mesh under the same settings (mesh size, starting point, probe count etc.) → Both should graphically give the same “bed shape” (use the same tools for visualizing, not Marlin Octoprint and Klipper Mainsail)
- Try to verify the real Z-Distance at some selected point, e.g. with a feeler gauge