Simple, it's called a selection bias. The people able to long-term control what they eat aren't obese, so by running a study on obese people, you're essentially biasing the study towards subjects who have difficulty controlling what they eat. This means most of the subjects will struggle to stick to the diet long term.
I think these studies controlled the diets entirely? The notion is that they did not deviate from the diet, yet their body adjusted to return to the original weight. Presumably by slowing the metabolism and decreasing energy for activity. If activity was controlled as well that would be truly fascinating. There is quite a bit of room for optimisation apparently. Through hikers often have trouble maintaining weight after stopping because their metabolism becomes very efficient.
