In this paper, we propose a new tool based on Observed Time Difference of Arrival (OTDOA) technique in order to estimate both the position of the different User Equipments (UEs) and the Relative Time Differences (RTDs) between Node Bs in real UMTS networks from measurements reports providing the TM parameter and the locations of the sites. This technique results in a non-linear least squares estimation problem which is solved by employing an iterative method based on the Levenberg-Marquardt (LM) algorithm and taking advantage of a particular and frequently found spatial geometry of sites: the star configuration. Modifications proposed in this paper increase its inherent accuracy while maintaining a reduced computational complexity, a fast convergence and a high robustness. Obtained results employing real data from various drive tests measured in different locations are compared with the performance of a standard LM algorithm run on these same measurements. We show that remarkably high geolocation and RTD accuracy is achieved considering only filtered events beforehand.