That mostly doesn't help, due to the fact that, as I understand the term, standard "decision theories" are intended as normative theories, meaning that when real behavior deviates, the investigators write the behavior off as wrong for failing to conform to theory, instead of trying to come up with a theory that actually describes the behavior.

So if you end up comparing an even-loosely-descriptive theory to a normative one, the descriptive theory will win, every time, no matter how far it is from true accuracy. That is, if you line up prospect theory, standard normative decision theory, and some new descriptive theory, the latter will definitely win in a model selection. But that doesn't tell us how the quantum descriptive theory compares to, for instance, the causal-probabilistic descriptive theory.

I don't understand your point in comparing this work to normative decision theories. The work is explicitly modeling experimental results that have been hard to model with existing probabilistic approaches.

That said, if we find that cognition can be well described by some fairly clean math, we should take a hard look at "normative" models that say it should be different. Up to now the descriptions have mostly had to use a pile of heuristics that could be (handwavingly) justified as "cheaper". But this work suggests a very different picture.