Also, the author completely missed the fact that different people react differently to different treatments.
It's basically akin to looking at Google Analytics and trying to predict what your next website visitor will do. You'll have an idea of what someone's going to do, but you can't perfectly predict it.
Also, people feel differently about different treatments, for particular reasons. If your mom always took one brand of Tylenol when she had a cold, taking it might reassure you more than the theoretically optimal painkiller. Customers were quite happy to pay more for the exact same Ibuprofen labelled with 'Back Pain' or 'Period Pain', because they felt it worked.
If you think about it, the author's desire to choose 'the scientifically proven to be the most effective' is just another example of such a superstition. If you could convince him that a particular medication was the winner in the meta-analysis, he would probably objectively feel better, even if it wasn't actually true.
It's basically akin to looking at Google Analytics and trying to predict what your next website visitor will do. You'll have an idea of what someone's going to do, but you can't perfectly predict it.