Study on LLMs for Promptagator-Style Dense Retriever Training