Epidemiological studies such as this one are inherently noisy due to the intrinsic presence of multiple factors and the lack of reliability of questionnaires. They should not be used to deliver ironclad conclusions, only to help construct questions that can be followed up on in more precise studies.
What are needed are metabolic ward studies involving a variety of diets, a large number of patients, and followed up over periods of at least 30 days, then comprehensive testing. Those studies are expensive relative to the available government grants, which is why we often see studies of 10 people followed up for 6 days, and other nonsense. However, they actually would not be expensive relative to a lot of pharmaceutical studies that can cost hundreds of million dollars, for drugs producing marginal effects.
Tofu can be good. How often do you go to Thai restaurants?