If 92% of the inmates who opted for InnerChange were drawn from among that non-reoffending 45 (say 12 inmates sign up, and 11 of them are drawn from that 45) then it does not indicate that InnerChange was a drtamatic success in the context of the whole prison population.
Even with the corrected math, there is not enough data given to prove the thesis that InnerChange had a dramatic effect overall.
The total reoffending rate among all inmates, not just InnerChange volunteers, is the key stat that is missing.
Unless I’m confused, the reoffend rate for all released prisoners is the 55% mentioned, with the reoffend rate among InnerChange participants at 8%.
This is, of course, a self-selected group. Presumably those who participated were those who wanted to turn their lives around and they would for this reason have been less likely to reoffend even had they not participated in the group. Those who are content with their lives as criminals aren’t likely to join the group.
But I still think the statistic is quite amazing.