A lot of scientific research work is done by underpaid graduate students who don't care too much about their particular research work, and wouldn't even really trust their own data. When you're cherry-picking studies where this is the case, one can get into a reality that's quite far from what's true.