I think dieting correctly is not so much a matter of discovering an ideal list of foods, as discovering the many things that can go wrong. Some of those things are more complex than any particular food or chemical being good or bad; in particular, timing matters (multiple small meals is better than one big meal), and variety matters (though I'm not sure whether it's good or bad). The goodness or badness of many foods fails to generalize between people due to genes, gut bacteria, and other factors.
There are a few things about diet that I am confident of, though, and paying attention to them keeps me looking good in my cloak. Sufficient protein is important. Carbohydrates and sugars with a high glycemic index (GI measures absorption rate) are bad. Micronutrients are important, but easy. Sufficient fat is important, and specific fat types seem to matter, but I don't know how exactly. Greatly overeating on special occasions is bad, and probably worse than people think. Noticing how particular foods make you feel shortly afterwards is important.
I'm less sure of this one, but I suspect that varying the number of calories consumed per day by too much is probably bad, and skipping meals or finishing a meal unsatisfied is probably bad. My reasoning there is that in the evolutionary environment, those would be strongly correlated with food shortage and should therefore trigger energy conservation.
Since your last paragraph is the idea behind intermittent fasting diets (and one hypothesized mechanism) I wouldn't be too quick to label it as bad: it's been linked, if inconclusively, to improved key blood markers and increased lifespan. It may be bad for weight loss – I don't know enough about nutrition to say, really – perhaps that's what you had in mind.
I've just read the Four-Hour Body by Timothy Ferriss. It seems on the face of it like ridiculously valuable material, if true - like what the completed version of Michael Vassar's proposed reboot of dietary science would look like at the finish point if dieting turned out to be more susceptible to Munchkinism than in my wildest dreams. Ferriss also talks the rationalist talk quite well in this book, much more so than in Four-Hour Workweek; he cites the experiments and occasionally says things like "I spent a lot of money on this and I expected it to work and it didn't work at all" or "I tried this and it seemed to work and I have no idea why it worked and I think it was probably a placebo effect."
Does the LessWrong hivemind have an opinion about 4HB? Has anyone tried it and found that it doesn't work, or that it does work, or that it works but not as well as Ferriss thinks it should work?