Can Large Language Models Recall Reference Location Like Humans?