See- you immediately went to a mechanism that justified the result. It's all too easy to convince yourself something is true because you can see a pathway- yet that pathway might not matter.
Take the analysis here on its face, instead: the same methodology (as much as can be reproduced - which is itself a huge problem) produces nonsense results in controls. The original paper did not have these controls.
The original paper is nonsense. It doesn't pass rudimentary bars for a scientific result, and should never have been published.