I suggest you measure some known objects - e.g., gage blocks, ring gages, even your calibration sphere, using the same probe configuration you're going to use to do measurements on your parts. The deviation from actual value(s) should give you a pretty good idea of what to expect in "real life".