There were 492 sex offenders tested, of which 63 were sent back to jail. That means 87% of those tested remained at liberty solely on the basis of the result of their polygraph. The reason given for poygraphs not being allowed as evidence or support for those maintaining innocence is that they are not reliable enough.
Whatever your thoughts on the test itself, the double standard here is quite clear, and has some disturbing implications - it's "reliable enough" to justify keeping convicted sex offenders at liberty, but our parole system is supposed to ensure that those who are released from prison have demonstrated a clear reduction in their risk of re-offending and, far more importantly, to be considered not a risk to the community to which they are returning. The use of these tests on sex offenders makes a clear statement that the authorities are not confident enough in their own decision making processes to accurately judge who is ... and isn't... a continuing risk to the community.
But those taking the deccision to leave the remaining 429 offenders at liberty have to be trusting that the polygraph results are accurate. On the basis of some claims (used to discredit polygraph results for those maintaining innocence) that the test is less than 60% accurate, 197 of those offenders could be at liberty not because they are safe to be free, but because their test results are wrong.
So which is it? The test is reliable enough that we can be confident that nearly 200 offenders are not dangerous (and therefore should be reliable enough to support claims of innocence), or it is not nearly reliable enough for any sort of confident conclusion (the excuse used for those maintaining innocence), in which case, we are actively putting people at risk of sexual attacks?
I'm not suggesting, and I don't believe anyone else is suggesting that a polygraph result alone should be evidence of guilt or innocence but, as with everything else, especially in so-called "circumstantial cases," they could be used to lend weight to other evidence.