Towards a Holistic Evaluation of LLMs on Factual Knowledge Recall