One of the more useful dynamic data used by ATIS (Advanced Traveller Information Systems) and ATMS (Advanced Traffic Management Systems) applications is the link travel time, in particular for the DRGS (Dynamic Route Guidance Systems) case. Among the different methods for collecting these data, the use of probe vehicles is one of the most efficient and promising method. However, accuracy issues have to be solved, above all, from the statistical point of view. This paper presents an analyse of the individual link travel time records distribution shape of a typical urban network link. The influence of this analysis on the probe vehicles-based estimation accuracy is also described. A division of the overall records in sub-samples according to the upstream and downstream link used by the vehicles is proposed. Advantages and disadvantages of this approach are commented. The necessity of integrating the link travel time estimation accuracy in a path context instead of treating it as an individual entity is demonstrated. Finally, some conclusions are given and further research needs are highlighted.