Can people change? Is learning possible? If so, then I think the more useful study is one where they examine the recommended use case. That way, we can test new methods of doing the same thing, and if we find a way that works better, people can adopt the better way! (I know, I know: wishful thinking.)
Your assertion, that researchers should only test "actual use conditions" (I didn't see the study where they surveyed people to find out these actual use conditions, but it must've been done, right?), carries the inherent co-assertion that people can't change, so we shouldn't bother studying new ways of doing things.
All this study shows is adherence to Murphy's Law: if people have the opportunity to do something the wrong way, they will.