Nearly all ratings are voluntary, and so suffer from self-selection bias. The measured ratings for general interest movie like Shawshank Redemption are typically lower than a special interest movie like Dark Knight (or Harry Potter, or Twilight, or Lord of the Rings) which appeals to a dedicated fanbase. The latter typically have a lot of fans who rate it highly just because it appeals to their group. That is, they rate it according to more lenient standard than they rate other movies, or they flat-out stuff the ballot box to try to get others to see it, to exaggerate the size of their interest group in hopes of encouraging more such movies to be made.
I agree that there is validity in this argument but a skew based on a fan base tends to be much greater when the number of responses is low. Dark Knight has over 1.5 Million ratings. Yes, a good chunk of them could be die hard fans, but there will also be offsetting low ratings by people who hate the genre, disliked the film, and/or people who didn't even watch the film. The law of averages kicks in to reduce the amount of skew as a result. Could it be skewed towards the high side because of rabid fans, possibly. The question is then by how much? and, does it even matter?
The biggest problem that I see is how people use ratings, as you aptly point out. They work well when being used to provide suggestions based on what you have watched and rated highly. The do not work well if your are just looking at the ratings and trying to figure out whether you will like the film or not. They also do not work well in determining the "best film ever", simply because that is way too subjective and is based on more criteria than just enjoyment and popularity.