Comparing Two Model Designs for Clinical Note Generation; Is an LLM a Useful Evaluator of Consistency?