Steve, that confidence level is based on statistical stuff that was well shown in a graph that's no longer available online. It's based on the simple fact that the more shots one puts in a group, the better it represents the accuracy one can expect for all the ammo with that load.
It parallels an old addage of measuring stuff that, in so many words says, if you don't get the same number every time you measure something, either your ruler is broken or technique is poor. So, if one shoots a few or several groups and they're not all the same size (within 5% for practical shooting accuracy work), they don't realistically define the accuracy of the ammo. The more shots per group there is, the higher ones confidence is it represents the accuracy all that batch of ammo has. Arsenals shoot several dozens of shots per test group with small arms ammo so their confidence level's well over 95%. They measure each shot hole's distance from group center then calculate the mean radius; about the best way to measure accuracy.
I may have saved that image and I'll look for it on my older computer then post it for all to see.
Last edited by Bart B.; October 10, 2012 at 10:36 AM.