Ok I am all out of ideas to test and looking for some more, I usually make more decorative stuff, but now have a project where all the parts need to fit together pretty accurately. I am sure I have had this problem since the beginning, it just has not bothered me much until now. I have parts 21.5" cutting on a 4'x4' table the right side of the table is cutting pretty much spot on but left side is cutting a 1/16" short. I have tested on .75 and .5 material, doubling the number of passes to see if backlash error was doubled, and I had the same error.
I have verified nothing is loose, nothing is binding, pinions show more wear than I would expect, but I really don't think it is a backlash issue (left side is cut first then the right, if backlash was the problem wouldn't it be more of a problem at the end of the program?). I have swapped drives, and changed my cable connectors (they were molex and were getting hot on the x and z, which means there was more resistance causing heat build up). For fun I have changed steps per to see if the error got smaller or larger and the error did not seem to change mine is calibrated over 45". I have tried mach3 and uccnc and it is the same, I remember one of the first projects I did was an end table and remember also being a 1/16th off but it went together, so I didn't worry about it much. I have not swapped motors yet, that was still on my list, and I was coming here to see if anyone else had some other tests I could make to start eliminating or verifying issues.
Electronics:
Leadshine M542 drivers
Nema 23 381 ozin motors https://www.automationtechnologiesinc.com/products-page/nema-23/nema23-381ozin-3-5a-single-shaft-stepper-motor-kl23h2100-35-4a/
UC100 (using UCCNC now with Gerry's screenset)
C35S BOB
Hitachi M12VC with Super PID (manually controlled not PC controlled)