Disambiguating Large Language Model Performance on the Ambiguities of Law, Reasoning, and the Future