"The guards were coached".

In a hierarchical system of "powerless and powerful", authority comes from the top, the low level enforcers are always trained, coached and must follow established standards (good or bad).

Not being relevant to what happens in the real world, removing context, was, apparently the study objective, at which they failed.

There has long been a disparagement of the so called soft sciences, like psychology and sociology, by those involved in the hard sciences, astronomy, physics, engineering. For example Neil deGrasse Tyson is famous for for his jibes against the soft sciences during his talks.  I usually discount most of the studies of humans by psychologists and sociologists. There's just too many things that influence people to get adequate control to one factor and you can't do to people what would be necessary to get those controls. Scientists can't do to animals what they do to inanimate objects or bacteria and viruses and can't do to humans what they do to lesser animals. In the end you're almost never sure what it is you've actually been testing for in the study.

This spot of bother makes me wonder if the experiment could have ever canceled out the influence of being an artificial condition. In the Stanislavsky teaching on acting, there is an insistence that one isn't making the set real but allowing a quality to reveal itself. The suspension of belief is entangled with recognition of something that is either there or not. It is a tricky bank shot where the problems of verification in theater get mixed up with isolating causes for behavior.
The "Boulder Model" developed by psychologists more than six decades ago was set up to establish forms of research that could bolster and extend the efforts made to help people by various methods. The original objections to it were that framing a problem to be the specific result of something was just like the language of the person seeking help. The standing wave of complete reflection.