[sesm]

Good point. In short, I would rephrase it as "if your measurement doesn't show the difference between trained and untrained teachers, maybe your measurement is wrong?"

There 2 other plausible hypothesis that the article ignores:

1. What if untrained teachers are effective because they learn on the job from trained teachers? Which means that you have to have certain % of trained teachers to keep the system alive.

2. What if trained teachers become less effective because they drop their standards looking at how untrained ones work? This is a common phenomenon at workplaces, and I've seen it happen in software development many times.