Text this: Comparative Evaluation of BM25–FAISS and Small-LLM–GPT in Retrieval-Augmented Generation Concept Map Assessment