Quality Digest

I have an issue that is becoming so convoluted I am getting confused, perhaps someone here can help me out. The trouble all began when we began manufacturing a ring gear which is large in diameter and thin walled, therefore creating a roundness problem. The tolerance on the bore ID was a total of 50 microns. We were seeing out of round condition of up to 40 micron, so therefore we thought we were in trouble, there was no way to hold 50 micron size tolerance is 40 micron is being used up for out of round. We got the customer to expand the tolerance to 100 micron total. The methods of collecting this data I also feel are somewhat flawed. At first we were collecting min and max diameters on a CMM. I know the CMM constructs these diameters by taking the largest and smallest points it gets. We also got a tallyround and with the tallyround are seeing roundness of about 17-20 micron. I feel if we show capability on the size, using an average computation rather than a min / max scenario, and then show good results on the tallyround we should be good. However the customer and other management seem to be held up on this min / max diameter business, I guess they feel if we are capable in a worst case scenario type of deal, then we are capable. However the print does not call out a runout or roundness, just a size. We only got into the roundess situation to begin with because we thougt the parts were too out of round to hold the size. Does anyone else feel this min/max method on the CMM is flawed, and if so what combonation of methods or method would be best to prove out this process?