Energy-efficient train operation is of high importance in urban railways. It takes into account both energy saving and punctuality at the same time, which is the goal of the most operators. In this study, the effect of employing variable regenerative energy recovery rate (RERR) for each inter-station was shown in energy-efficient operation improvement. For this purpose, a two-stage optimisation was proposed. In the first stage, which is a mechanical optimisation, optimal speed profiles for a single train were determined by a bi-objective optimisation through a non-dominated sorting algorithm. In this stage, a simulation model for determination of train's performance was developed and utilised. It was shown that variable RERRs eventually led to different energy-time Pareto fronts. In the second stage, the total input energy for a multi-train system was minimised by using obtained optimal speed profiles from the first stage and distributing total travelling time among inter-stations (electrical/electromechanical optimisation). It was shown that optimisation of the net energy – with different values of RERR – instead of consumed energy, can reduce the total input energy of the network. The simulation results, which are based on actual operating data of Mashhad urban railway system, confirm the feasibility and effectiveness of the proposed method.