One of the issues is that by chance the models used have an unexpected behavior that seems to confirm the initial hypothesis is sound. And this is not rare. So

One of the issues is that by chance the models used have an unexpected behavior that seems to confirm the initial hypothesis is sound. And this is not rare. Sometimes it is fraudulent, people know things are faulty but continue anyway. In other cases it could be by ignorance of the existence of the phenomenon behind the faulty behavior or simply because that phenomenon has not been discovered yet.