I don't agree that this is reversion to the mean. If we saw it be the case that school districts with particularly good outcomes one quarter were more likely to

I don't agree that this is reversion to the mean. If we saw it be the case that school districts with particularly good outcomes one quarter were more likely to have a decline in performance the next quarter independent of all observables, THAT would be reversion to the mean. If there is a statistically significant difference in outcomes between charter schools and non-charter then I don't think we can dismiss it on the grounds you state. But you did bring up an excellent point in the self-selection bias. We are potentially comparing apples and oranges if the charter schools get a different sort of student to teach.