> then it should be easy to author your own paper that proves whatever is "obviously true". Or maybe reality isn't so easy?

Yes, this is my point. It's very easy to publish papers which say anything, if you don't mind that the results are junk or fraudulent. This is why most science is un-replicable, and you should have low confidence in a randomly selected paper, especially if the results conflict with bayesian expectations.

It's actually very hard to perform and publish good science.