I guess it is mostly the cases that have previously been screened for being likely overturning candidates that are actually brought to court to get overturned.
There's no particular reason to think this is true. Availability of DNA evidence or eyewitness evidence is relatively independent. Thus, it is reasonable to treat the DNA & eyewitness (DNA+e) cases as representative of all eyewitness cases.
In the DNA+e cases where DNA is inconsistent with guilt, either the DNA or the eyewitness must be wrong. And we have independent reasons to think DNA is more reliable than eyewitnesses. If DNA really is a representative sample of cases, and DNA+e is a representative sample of DNA cases, the wikipedia statistic you cited suggests that 75% of eyewitness testimony is wrong.
As an aside, the research is pretty clear that there is a substantial difference in eyewitness accuracy based on whether or not the witness knew the perpetrator.
Now, there are reasons to believe the DNA cases are not a random sample of crime, particularly because certain kinds of crimes are more likely to leave analyzable DNA (eg. rape vs. bank robbery). But that doesn't suggest that they aren't representative of eyewitness cases.
But at least the court outcome will skew the selection.
Also we have plenty good statistics for eye witness reliability when it comes to UFO sightings and these statistics give a very different conclusion:
"Only 1.5% of all cases were judged to be psychological or "crackpot" cases" http://en.wikipedia.org/wiki/Project_Blue_Book#Project_Blue_Book_Special_Report_No._14
Shouldn't we use the most relevant statistics?
Recently I've been struck with a belief in Aliens being present on this Earth. It happened after I watched this documenary (and subsequently several others). My feeling of belief is not particular interesting in itself - I could be lunatic or otherwise psychological dysfunctional. What I'm interested in knowing is to what extend other people, who consider themselves rationalists, feel belief in the existence of aliens on this earth, after watching this documentary. Is anyone willing to try and watch it and then report back?
Another question arising in this matter is how to treat evidence of extraordinary things. Should one require 'extraordinary evidence for extraordinary claims'? I somehow feel that this notion is misguided - it discriminates evidence prior to observation. That is not the right time to start discriminating. At most we should ascribe a prior probability of zero and then do some Bayesian updating to get a posterior. Hmm, if no one has seen a black swan and some bayesian thinking person then sees a black swan a) in the distance or b) up front, what will his a posterior probability of the existence of black swans then be?