It's on point because "number of students who can integrate in 2012 vs 2011" is rather more like "percent of regression tests passed in 2012 projects vs 2011 projects" than it is like "quality of (?) search engine code vs quantum simulation code", or better, "number of successfully completed customer tasks using search engine product vs quantum simulation product". That is it measures something that is not irrelevant, but awfully specific, and possibly less indicative of the larger whole than you might think. Because teaching is pretty heterogeneous. And the measurement is extremely indirect, i.e. how good another person is at something after interacting with the programmer/teacher's product.
