Freeform RPS alignement recalculation help

[[ED...]]

Hello,

Recently I´ve been struggling with the inspection of a checking fixture specially with the RPS alignment. For this device we have 6 spheres which we use for the RPS alignement as follows: all 6 spheres on X axis to create the plane, 4 alligned spheres on Y axis to create a line for the 2nd plane and 2 spheres (alligned between them but not with the previos 4) for the Z axis and of course since Im not following the 321 rule of RPS alignemnt I have to use the Freeform RPS to match all spheres
The result is that all 6 spheres are out of tolerance which invalidates any measure of the device since the RPS alignment is wrong. Because I dont know if the alignment is wrong or the spheres are out of position, I go to "change current base alignment" and replace the nominal coordinates with the actual coordinates obtained from the previous alignment with the expectation of getting 0 deviation and compare the colorimetry of both results to see which one matches better the scanned point cloud to the CAD data.

The problem is that after doing this, the deviations are not recalculated so I get the same deviations from previous alignemnt (in the pictures shared I only slightly modified Y and Z coordinates); does this means I cannot update a RPS alignement and I need to start a new measurement plan if I want to change the coordinates to evaluate a different alignment? As a side question, is my original alignment rightor should I stick to 321 RPS and choose only the spheres that are closest to their nominals?

Thanks in advance for any help anyone can provide

[[Me...]]

I think you need to make the best fit in your alignment
baseSystem().valueA<0.005

RPS_Alignment (1).pdf

[[ED...]]

I´ve tried, I get similar results but the problem is that I get too much variance between each alignment test, sometimes as big as >1 mm (with a tolerance of +/- 0.33 mm), so I cannot define a coordinate without getting an out of tolerance flag again in the next run of the program.

I tried the RPS freeform alignment of the spheres with a loop
I tried RPS freeform alignment of the spheres and then a bestfit (even eliminating everything from the CAD except the spheres and the aluminum plate base to avoid any influence from the other bodies on the bestfit)
I tried bestfit alignment with the alumimun plate base elements and then obtaining the position of the spheres
And finally a bestfit alignment with the main body and then obtaining the position of the spheres
This demonstrates that spheres are not within tolerance of their nominal position so I guess my true question is: if I use the coordinates obtained from all my tests (and not the nominal coordinates from the CAD) to get the spheres within tolerance, is it a valid alignment?