Candiru, as interesting as your link is, I'm not convinced that the methods described apply here. Each shot in a group can be used to calculate a value such as "average distance to center". This is shown as ATC on the targets in my earlier post. In this case, each shot is statistically significant and in a sense carries equal "value". However, if in a 20-shot group, 19/20 shots fall within 1/2 MOA, but one shot is out at 1 MOA, the group size is 1 MOA. In other words, group size is not a statistical average because the one shot out of the group (flyer) doesn't carry equal value.
Consider my example in the earlier post of the 1/4 MOA rifle, 1/4 MOA load and 1/4 MOA shooter. Assuming all three components perform at 1/4 MOA, a shot could be to the right of the POA by 1/8 MOA (due to the shooter) + 1/8 MOA (due to the barrel) + 1/8 MOA (due to the load). If a shot to the right at + 3/8 MOA is possible (max error to the right), then a shot to the left by the same amount is also possible (max error to the left). This would mean that if all three systems perform to their respective accuracies, all shots should statistically fall within 3/4 MOA. Naturally, there'll be a Gaussian distribution around the center of the target and there will be groups smaller than 3/4 MOA, but the system should consistently be a 3/4 MOA system.
I'm curious as to what others think about this. I could be way off.