1. This kind of research is difficult to execute because there are too many confounding variables. You can never get clear isolatable results. There is no such thing as "good" and "bad" that applies to a large population in all contexts. This may sound like a drawback, but in a way this is great for researchers because it creates an infinite amount of studies you can run and publish, and nobody can ever tell you that you're wrong, because there is no clear answer. You'll never be out of work. If industry tells you "we need a study that shows X" there's almost certainly a way to design a study in such a way to get result X.
2. Research must be published. No matter what your results or experimental assumptions and constraints are, as a researcher you must "sell" [0] your research and present it in the best light. That's what your career is all about. Citations. Similarly, as a reporter you must write articles that attract views, so you must sensationalize research, even if the original publication was clear about its shortcomings. These perverse incentives create vicious cycles, not just in nutrition.
[0] https://www.cs.virginia.edu/~robins/YouAndYourResearch.html