Most existing evacuation behavior simulation models for large-scale urban fires following major earthquakes rely on extreme idealizations and simplifications. These models often assume that residents begin evacuating immediately after the disaster and follow the shortest possible route. In this study, we develop a more realistic simulation framework by incorporating models for evacuation start timing and route selection, based on questionnaire survey responses. Additionally, we enhance the simulation by integrating key factors that influence evacuation behavior, such as the tendency to follow other evacuees and the means of acquiring disaster-related information. Using this improved model, we analyze evacuation strategies for high-density wooden housing areas during a major earthquake. The results indicate that in regions with a high risk of fire spread, direct evacuation may be more effective than a two-step evacuation process. Furthermore, the presence of early evacuees acting as influencers can significantly impact overall casualty numbers. Therefore, establishing a system for disseminating accurate and timely disaster information is crucial for improving evacuation outcomes.