I recently did a quick comparison between the automatic calibration and a laser measurement of the anchor positions (link to test). From what I can see, lower calibration current threshold seems to improve accuracy — possibly because of reduced flex or stretch in the system?
As discussed in this thread, belt strech might be a significant contibutor to overal machine accuracy.
If it’s of any interest: I have access to a universal testing machine with an extensometer. If there’s no official data from the vendor, I could try to measure the belt stiffness — provided I can get a piece of spare belt.
Would that be useful to anyone?