The approach sounds good, but the question is: how do we know it is good, and how good?

From what to 98%? These numbers are doable even in tough areas (Harlem Children's Zone is in that ballpark), but it's rare. None of the four links you gave make it easy to see outcome metrics, which is what I think you were responding to in the first place. I would prefer data over glowing testimonials. None of the four links seem to provide accessible data (before/after graduation rates or anything like that).

If it works anywhere near as well as you claim, documenting that it works well should have been easy, so presumably such documentation is around somewhere where it could be referenced.