Exit Polls Unmasked?

A colleague pointed out this morning that the exit polls for some races seemed to be off the mark this year. Like many people, I had the impression that exit polls had previously been very accurate but had gone awry in 2004. It turns out that this isn't necessarily so, according to an interview at Neural Gourmet. This paper (by the same political scientist) attempts to rebut arguments that the 2004 exit polls were evidence of fraud. Wiki has a good overview of the 2004 exit poll debate. (By the way, an intriguing new campaign trick is to tamper with your opponent's Wiki bio. Shades of Nixon!) Query: Does the possible unreliability of exit polls justify suppressing them until the polls close, or is this another area where the best cure for speech is more speech?