using .2 and .3 probes for the first time

The changes I made to match the settings picture Kai sent did not help. I have to check another part for the afternoon and wont be working on the problem part tell tomorrow morning. I did some experimenting with the help of my fellow CMM programmer and it seems like the .2mm probe is off in the x only the z and y seem to be ok. It leads me to believe its calibration but i can think of anything I did wrong. When I come in im going to reset al the tips. I'm thinking of deleting and remaking the .2mm probe system just incase iv got a corrupt probe file.

Its a new program but I have run several parts through it before this problem came up so I have a hard time thinking its the program. Any thoughts or ideas are certainly appreciated.

my apologies, I misunderstood.
--so you are measuring a deformed cylinder wall (the only way a spherical ball isn't falling out, is if someone swaged/pinged the cylinder walls after inserting the ball, yes)?

This is measuring a post-processed feature. Never a good practice.

If you are expected to verify that the diameter of the deformation is sufficient to retain the ball... You will never nail this correctly with a CMM.
There's just no way you can control measuring the high points of that swaged cylinder wall.

I would recommend either verifying this diameter prior to the ball being inserted (if you can)
-or-
Making a functional go/nogo gage, with a concave sphere at the tips (concave to clear the bearing's max diameter). You can even use this as an offset pin to locate the hole if that's your necessary goal (and not just size). ​

Another option is to use vision to measure this diameter. Comparator or VMM with some oblique-angle lighting should be capable of extracting that bore diameter after being deformed.

The ball bearings are press fit. But correct the holes most definitely are now deformed from assembling process. The issue is now that the product is for calibration of some sort of surgical robot. They need the exact location of the ball bearings on the part for calibration. What I have been doing is measuring the bores to get them the X,Z locations, then scanning over the center of the ball bearing in relation to the bore to get them the highpoint of the ball bearing. This gets them their Y location. I have been saying everything you said above about the repeatability of checking these small features. The product is to big for our vision systems, but I'm beginning to wonder if it may be worth sending this to hexagon with their big vision systems.

I recalibrated, reset all the tips before doing so. Seems to be running OK. We are a small shop in a pretty rough building. We are moving soon to a bigger better building. Could temperature fluctuation be the problem? We do our best to control temp. but its certainly not perfect.

I don't know how big those balls are, but I can't help but think you might be best off using a larger probe and measuring the balls themselves as spheres.

Even if you can only measure a small surface area of each ball, you can reduce a lot of measuring error by assuming the diameter of the balls to be nearly perfect. You can measure the balls directly as spheres using a FIXED_RAD math type. That should get you relatively accurate ball position measurements even if you can only measure a small arc of each ball.

Keep in mind that even a pretty average Grade 25 ball has a diameter tolerance of +/-.0025mm. So, it would be pretty safe to assume those balls are at nominal size. If you want to be even more precise, you can find the certs for the lot of balls that were used. That should tell you the average ball size with even more confidence. For example, Grade 25 balls have a lot diameter variation of only .0006mm. You can then use the average size reported on the cert for your fixed radius measurements to make even more precise location measurements.

TCOE is really insignificant when you have smaller features.
Vibration: (forklifts, machining operations, trucks, air compressors, etc) can be a major factor as well.

The real issue is reproducibility of a feature that has been subjected to post-processing. Pressing that ball into the cylinder is absolutely misshaping that cylinder wall, and your ability to have part A measure similar to part B relies heavily on how that cylinder wall was (or was not) damaged during that process. I bet you can see -with the naked eye- rifling or gouge marks around the cylinder walls yes?

How does one meet dimensional or locational controls of a diameter that is eccentric?!

I highly recommend getting pin gages made with a concave end cut into it. Simulate the depth and location using the pins.
if you know pin is 1.002" long from the theoretical centroid of the concave end, just measure the pin diameter as you do the walls of the cylinder (as close to part face as you can to minimize projected error) to make X Z.
Then, measure the end of the pin for your Y including whatever known offset the pin's overall length introduces along Y.

Chris that's a good idea. Ill try that tomorrow. Thank you to everyone for the help. I'm hoping that after this initial run the customer will allow us to report the location of the holes then instead of after the ball bearing being pressed in.