I wouldn't have said either of those sets of numbers was enough different that it mattered which you picked. It the numbers were a bit further apart, then it could make some difference.
Good point. I have certainly used this approach with words that seemed similar but had radically different degrees of frequency. Having said that, it's also a 'psychological hack' in that it helps me make a decision over which word to use. Somehow, knowing that one word is used even slightly more often than another seems to give me more confidence in using that particular word than if I simply flipped a coin... unless, of course, the number of searches is the same for each word! (Uh-oh, gotta think for myself again...)