Finding: Accuracy will never reach 100% because, regardless of model size, search and reasoning capabilities, some real-world questions are inherently unanswerable.

Translation: PEBKAC. You asked the wrong question.