I confess that I didn't read the article, but this reminds me of the HN story of the University professor who asked students who cheated to come in to his office, and then found some large percentage who came to the office though the software exonerated them. The question is: did they cheat and get it past the software? Or were they, as suggested in this article, admitting to something they did not do?
I suspect that there are similar factors involved in other false confessions. It's an asymmetric knowledge game. The only winning move for the person being questioned is not to play, but they may not be aware of that - and worse, they may trust the other player. With further intimidation, privation, and humiliation, it's not hard to imagine why people could break down.