It really depends on what you're trying to find and the data set. In any situation, the more samples you can take, the better. You may find that 7 samples in a given situation gave you the same statistical significance as 10 samples for a single test or data set. But again, you're only talking about increasing your sample size by 3. It's not so much 7 vs. 10 being the issue, but only having taken 3 more samples. Unless one of those 3 were a serious outlier, you would always find that 3 more samples isn't going to give you a stronger result. Saying 7 is better is relative. Those extra 3 samples would be negligible if they were Samples 8, 9, and 10, or if they were samples 101, 102, and 103. And obviously 100 samples would be better than 10, but I wouldn't say 7 is better than 10, only because they are so low a sample size anyway.
5 samples/shots seems like a good test to narrow in, but I'd like to see a follow up test from someone with the capability to cut and re-crown barrels. Take the top 5 or 6 barrel lengths and run 50 to 100 shots per barrel. With a 50 round box of shells in my area running about $14, the total cost in shells would be less than what you would make selling a single rifle. Meaning, if you got just one more person to buy a rifle based on the data, it already paid for itself.
The problem that I've always seen with tests like this were mentioned two posts beforehand. Different barrels, different shooters.....all factors will produce different results. When you add things like rate of twist, bullet stabilization, and other environmental factors to the mix.....this type of study is highly subjective.