The funny thing about recruiting rankings, and why I think they have some predictive ability, is not because I think the scouts hired by these outlets (Scout, Rivals, ESPN, etc.) are able to accurately assess talent in a couple viewings at a showcase. It's because the recruiting rankings are trying to predict likely offers. So, a well-connected analyst for one of these websites sees that a kid is getting strong mid-major offers and when he sees that kid in AAU, he's going to be anchored to a 3-star unless the kid is awful or fantastic. Much the same, if a kid has limited offers from low, low D1 teams, the kid may struggle to even get a rating and surely not one above a 1- or 2-star.
From my observation, offers tend to be better correlated with outcomes on an individual level, so my presumption on why the recruiting ratings have some predictive ability has more to do with the scouts' attempt to predict (or explicitly rate based on) offers extended rather than their assessment of the prospect's future D1 ability.