Summary:

Off-center the beam on curved mirrors significantly in one direction and everything is within tolerance (first plot, this is the best quality data we managed to squeeze out of this setup).

Center the beam on the curved mirrors and the data looks OK-ish but things are somewhat out of nominal tolerance (second plot, some green error bars are outside of two vertical lines representing the nominal tolerance), and the ellipticity and the astigmatism become worse (you can tell by the fact that X data and Y data moved in the opposite direction).

The third/4th attachment show how off-centered the beams were on the secondary and on the primary when the good data was obtained.

The 5th/6th attachment show the centering on the secondary and the primary when the second set of data was taken.

This is repeatable. Every time a good looking data is obtained, the beam position on the primary is offset to the left. Every time we re-centered the beam on the primary and the secondary, the scan data is worse (but still OK-ish, not terrible).

Sounds like a problem of the mirror surface figure to me. Maybe we got unlucky on this pair.

This is the best we can achieve, and since it's not terrible when the beam is centered on the curved mirror, I'll declare that this is the final tuning of the H1TMSX. Tomorrow we'll mate the ISC table to the tele.