"The reader is assumed to be an intelligent rational agent. That's a very poor assumption. But even if it weren't, it would, on its own, be insufficient to justify the practice."
It's not a poor assumption. It's a good assuption. In fact, prior training in the matter is necessary in order to read and comprehend the material. the practice is entirely justified, because to do otherwise would over complicate any work with an extreme amount of redundant, unwarranted comment. Where appropriate, footnotes are given.
No it isn't. Yes people can be intelligent and rational but also unintelligent and irrational. Further, the target audience of popular science writing very often don't have the necessary training or knowledge to distinguish claims that are well supported from those that aren't.