Hempy, this is a good illustration of a piss poor argument (on your part) and the sterling example of how one person can defy his captors. (Stockdale was another sterling example).
What you might wish to remember is that each person is a different case, and you don't know what he'll do or not do until he is tested. Hell, he probably didn't know what he was capable of until he met that test.
The above doesn't mean that one should go on a default mode of "try and see" with the abuse. Not hardly, the narrow view of believing that "push button pain produces info" (like a vending machine) has been shown to be a bad basic assumption. Yet it sometimes works. Therefore other considerations have to be added if one is to arrive at actual criteria to base such a decision on. Policy, image, risk, what we do and don't do, and why ... all of that factors in.
This post is to remind you that oversimplifying the assessment based on an example of heroic defiance is a careless use of the halo effect.
Put another way, not every soldier is Alvin York.