His analysis doesn't depend on how many significant digits there are: he's doing a hypothesis test, so all that matters is whether one number (a p-value which was not shown) is bigger than a threshold (.01). The fact that it is in most of the cases means that the variance in results that you mention has been handled by the sample size.

So I do think the results are meaningful.