No it doesn't. Your logic here is faulty. If the error in a given rating is random, then aggregating the statistics helps to remove the error. Looking at 5 statements could give you a distorted picture, but looking at 50 or better yet 500 statements will give a you a pretty accurate picture.
Now of course, if the error is not random, that is a different story. But that is where deeper analysis is needed. You can't just say, "Oh Politifact is biased and their ratings are garbage!" If the bias has a noticeable impact on their ratings in one direction, than that would be demonstrable and you would be able to show us a case, backed by evidence, that backs up your assertion. If they are writing solid and thorough analyses, as you concede they are, then chances are they assign error in an honest way - i.e., in a way that attempts to be fair to the person whose statements they rate. Making it likely that your disagreement with their ratings is merely that, well, you disagree with them sometimes.
But that in no way invalidates their use as at the very least, a blunt tool with which to assess who tends to say more factual things.