At the Discovering Biology in a Digital World blog, there's a post that questions the accuracy of DNA sequence reads. The author does a short experiment, calling the same short snippet of sequence once using phred and again using the ABI KB base calling program, but comes up with totally different graphs. Considering, as she writes in a previous post, the need for absolutely accurate data when it comes to, for instance, predicting disease predisposition, how can you be sure what you're looking at is a quality sequence?