It's possible to investigate whether diets produce beneficial results in subgroups even without having a proper general model of diet and nutrition.

It's analogous to investigating whether drugs produce beneficial results in people with X disease even without having a proper model of how exactly X disease works.