CAA probe calibration magic

[[Be...]]

I recently had a problem that took me way too long to figure out.  I had a part that measured great except for the few features that were measured with the angled probe.  I beat my head against it and eventually discovered that the issue was the calibration with the regular straight 3mm probe on a 75 mm stem.  It was off by almost 0.03".  I think that what had happened was that somebody had not revalued the reference sphere before calibration.  The interesting thing is that the numbers on the calibration were great.  It looked like a good calibration. Good sigma etc.

With roughly 100 features measured using over 10 probe orientations of basically every approach angle with that probe it agreed with itself.  Holes on one side of the part were directly in line with holes on the other side of the part.  I would expect that to not be the case.  The Y direction was off by almost 0.03" (not to mention the issues in X though Z was good). So when I am probing from the +X side vs probing from the -X side with a generic probe, if the reference sphere location was wrong by 0.03" I would expect the holes on the two sides to be off by 0.06" relative to eachother.  

I think the CAA just does a ton of magic to bring those into alignment so that anything with that probe is basically good to other things with that same probe.  But if you use another probe with a regular calibration ... especially a good calibration ... the two probes will disagree by the offset in the reference sphere location, or about 0.03" in this case.

Does that sound right?  Does anybody else have more insight or experience with this?  I cannot afford the cmm time to truly test the theory.