Can you think of any group this would not be true for given a sample size or population as large as the military?
Honest, decent people who join the military (or any other group) will likely remain honest and decent. Dishonest people will most likely remain dishonest until their actions prevent them from getting what they want.
Perhaps those values from military doesn't have/had any resonance because they weren't, in fact, good? Or, told in a good way? (I'm thinking now that tyranny in teaching generates repulse not acceptance)
Also, if we consider that the teaching was correct (both as content and form - I dunno of course) who knows if the "soldiers who didn't change" would be worse if this teaching would be applied?
I think that the human being is the victim of influence and good influences play a determinant role on his behavior. But now what means "good" - this is entirely another chapter...
Secondly, I don't think the ideals help if you don't agree with them. You choose to become an Doctor and remember the Hippocratic oath. Same with Engineers and Lawyers. If you find half way through, that you don't want to be an Engineer or a Doctor, you can always bail out and switch. But with the military, bailing out is completely out of the question.
While people may or may not like what the military does, most of its creeds/values are for the most part positive things. The Army values for example are Duty, Loyalty, Selfless Service, Respect, Honor, Integrity, and Personal Courage - pretty generic and not nefarious at all.
But since you asked, why would one have to have experience with that exact group of people prior to coming to a conclusion? I agree that it would be ideal if that was the case, but there are many studies considered to be scientifically valid that use one group of people who come from similar circumstances as a control, while only testing on another group. In my example I suppose the control would have been the people I know from outside the military. As I stated though, my comment was just an anecdote from my personal experiences, not a peer-reviewed, published article intended to expose the author of the other study as a fraud.