Because I could see a lot of variability amongst those. And without controls, the study will default strongly toward fast-food, doughnut shops, pizza, burger / franks stands, and the like. Several of which have pronounced associated negatives (see, e.g., Morgan Spurlock's Super Size Me <https://en.wikipedia.org/wiki/Super_Size_Me (2004).
I find some restaurant meals strike me as far better than others. Taquerias, Thai and Vietnamese food, better vegetarian restaurants (say, Greens in SF), for example. Specific choices such as sides, beverages, alcohol, and whether or not the restaurant permits smoking (some parts of the US still allow this barbaric practice) would likely be huge confounding factors.
I'm not discounting home-cooked meals, and generally far prefer them myself. But overly-broad, undifferentiated analysis is ... not especially illuminating.