LLMs Can See the Smoke but not the Fire: Evaluating Abductive Reasoning with Elenchos