You gererate a hypothesis and then experimentally test that hypothesis, I don't see anything wrong there. How else would you evaluate new theories that go against convention? Maybe you are suggesting that he fudged the data or otherwise manipulated the experiment to get a desired result?
Of course, but there is a difference between having a hypothesis to test, and believing something and setting out to find evidence for it. Those are some extreme examples of that being an issue, but also things like not publishing negative results can also be a result, less obviously.
I very specifically said that I was not saying that was what happened in this case, just that I agreed that there was something of a red flag.
One way that you can get bad results is to start with too broad a hypothesis, like 'autism is related to maleness', conduct your experiment and then engage in p-hacking, or analysing all the relationships in the data that would support your hypothesis until you find one that looks significant. [https://bitesizebio.com/31497/guilty-p-hacking/]