Most of our current long-term (>1 year long) studies about diets deal with the problem of "subjects not sticking to the diet" by just saying that X% of participants stopped the diet and then proceeding to analyze the health benefits for the Y% that stuck to it.

That's a fair thing to do, but what's not supported from those results is the conclusion that such a diet would have worked very well for the X% had they stuck to it.