Bus rapid transit (BRT) systems are a cost-effective public transportation solution moving millions of passengers every day. To optimize their operation, it is usual to implement macroscopic models that neglect the microscopic details of bus motion. In this work we show that the docking bay assignment (DBA) for the bus services at the stations, one of those disregarded microscopic details, has a large impact on the overall BRT performance. To evaluate the variations in the system's performance upon changes on the DBA, we have simulated the entire operation of a simplified BRT system using a cellular automaton microsimulation scheme. We find that the critical service frequency, above which bus queues appear, strongly depends on the DBA at the busiest stations in the system, leading to significant performance differences. Smaller critical frequencies, and poorer performance, are observed as the service demand for the services sharing a docking bay increases. By approximating the frequency optimization problem, we show that, due to the limitations imposed to the optimal frequencies by the critical frequency, the service frequencies that minimize the total cost depend on the DBA at the busiest stations. This correlation becomes stronger as the passenger demand increases and as more importance is given to the passenger time in the total cost. Our results suggest that the frequency optimization problem in BRT systems must include the docking bay assignment at the busiest stations as an additional operation variable.