Decentralized synchronization among peer devices (PDs) for distributed device-to-device (D2D) wireless networks has been explored by many researches. A general way of performing decentralized synchronization is based on pulse-coupled methods. In which, all PDs are equal and each PD sends and receives synchronization packets as well as adjusts own clock accordingly to achieve mutual synchronization. A newly published standard of IEEE STD 802.15.8 (STD-15.8) provides a systematic process to achieve synchronization for distributed D2D. However, we find that the synchronization process of STD-15.8 may results in separated synchronization clusters. To solve this problem, we propose additional procedures to that of STD-15.8. Then, we conduct step-by-step simulations to reflect the lower MAC functions of real transceivers by employing multiagent model. The simulation results verify the effectiveness of our proposal on improving synchronization.