When you look for a periodicity in a bunch of data, you might do a fourier transform and look for peaks. (The other work did that.) Essentially you're doing a large number of correlations. If you try 100 corelations, you should not be surprised to see one that's significant at the 99% level, just by chance!
On the other hand, if the periodicity they found matched known astronomical phenomena (as opposed to the other work I looked at), the probability of a chance result is much reduced.