Instead of looking to see what other people/machines are getting you should find out what your machine and probe configuration produces. You should have this data available and documented should there ever be any questions arise concerning your method of determining your actual error.

You need to set up a "standard" measurement using the best probe configuration (short and small diameter) in CNC mode. Set XYZ zero on your calibration sphere or some other certified artifact and measure the artifact 10 times recording everything- XYZ locations and size. Next, set up to repeat the same measurement process using the long probe and diameter that you plan to use, measure and record everything. Compare individual and group averages. If you feel the error is too great try another probe set up until you are satisfied. You will need to have an error of less than 10% of the feature's tolerance to pass a critical auditor's questions.