I have been using MEGA6 for a while in order to evaluate some stats of my gene of interest.
One of the tests that is possible to conduct with MEGA6 is diversifyng/purifying selection. So here is the question. If I try to run the test with two divergent sequences on several exons of a gene I can find in many cases that puryfing selection is occurring (dS > dN significant). However, if I run the same test on an intronic sequence I also get a positive signal of purifying selection in many many (too many) cases. While there might be some selective pressure also acting on introns it is impossible that this is acting on a codon-like structure. I don't think that the issue is in the software so I guess there must be some flaw in my argument. However I don't see how a good statistical test could reject the null hypothesis if clearly random sequences (with respect to the test) are used.