Intervention vs. Control Group for Matched Pair, dichotomous data: McNemar Test?

I have matched-pair, pass/fail data on a reading test for 2010 and 2012. I have this for an intervention group and a control group of students. I'm trying to show that my intervention group did better than control group. I've calculated two McNemar statistics for both the intervention (sig.) and control group (n.s.) Now I'm trying to compare the two groups. How do I compare the McNemar stats? Or should I be running another test? Thank you in advance.
This kind of data is messy because the ones in each group who passed at baseline can't get better -- they can only stay the same or get worse. Opposite for the ones who failed at baseline. So what is the real critical question?

If both groups had roughly the same proportion of failures at pre-test, maybe you can just take those subjects and compare the proportion who improved or didn't improve between the two groups. So, of 20 baseline failures in group 1, 10 improved. Of 20 baseline failures in group 2, 18 improved. Use chi-square to compare them.

Then, to be sure your intervention isn't hurting the good students, also compare the ones who succeeded initially to see if one group more often regresses!

If one group had many more pre-test failures than the other, it may tend to improve more just because some of those many failures are accidental drops below ability that will self-correct (regression to the mean).
Thanks. Is it wrong to eliminate the students who didn't change from 2010 to 2012 (passed/passed or failed/failed) and then compare the remaining students. So say something like: of the students who made a change (passed then failed or vice versa), 40% of the control group failed and then passed vs. 60% of the intervention group who failed and then passed, chi square=...