Kossel Linear Plus Dimension error

4chs

Hello.
I have Anycubic Kossel Linear Plus. Week ago i switched to Repetier. After that, i calibrated whole table to be flat using Escher3 wizard. Just perfect with default Rod Length 271.5. (Real ~ 267.8. But with this value, i have problem with opposite points. On X, Y, Z table = 0mm, on XY, YZ, ZX = ~-1.5-2mm)
But my prints are smaller, than it must be.I printed this: https://www.thingiverse.com/thing:1274733. It must be 75mm, but i got ~74mm.
From this point, things became weird. If i decrease rod length to suggested, i have problem with opposite points. If i calibrate table to be flat, my prints are smaller.
What should i do with this? Any advices?

Repetier

There are 2 escher solutions. use the one with fixed rod length and set it to known length. If it is 4mm off your geometry will be wrong in size. That is something z probe does not show as it can not test and your optimization diverges from real geometry just to optimize z.

4chs

But with "Real" rod length, my table looks like Z 0.0; ZY -0.2; Y 0.0; YX -0.2; X 0.0; XZ -0.2.
The only way (as far as i know) to rise opposite poits is to increase the rod length.
That is really strange.

Repetier

The question is here how good can you measure? You can correct in the range of measure tolerance of course, but 4mm is more then tolerance. But remember to measure from rotation center to rotation center! Horizontal radius might also have influence on this but then the center also changes height.