Okay, but there are basically no studies showing no or negative effects. Except for ones with massive p-values. You'd think there would at least be a few showing no effect.
* A lot of small trials that are perhaps not even statically significant and show some positive effect.
* A few very large trials that show essentially no effect.
Very small trials are easy to organize, and if it fails people gets disappointed and forget to write it down, or the journal reject it because it shows (almost) no effect. So there is a lot of filtration.
Very large trials are difficult to organize, so they must publish the final result somewhere to ensure they can fill all paperwork. Also it's large enough to be interesting for the journal. So there is less filtration.