When Truth Matters - Addressing Pragmatic Categories in Natural Language Inference (NLI) by Large Language Models (LLMs)