A Bluetooth traffic monitoring system (BTMS) is capable of identifying vehicles and estimate their travel time (TT) in a route. This information is key for intelligent transportation systems. Although BTMSs are currently deployed in several cities throughout the world, there is no formal methodology for the TT estimation they generate. In this paper, we first analyze the specific features of the Bluetooth technology that affect the TT estimation. In particular, we study the reliability of the measurements, the representativeness of the estimates, and the issues regarding multiple detections and outliers. Based on this knowledge, we propose a comprehensive methodology for the TT estimation that considers exclusively information from vehicles. We filter these vehicles through a simple process that uses the available dedicated inquiry access code. In order to illustrate our proposal, we performed an experiment deploying commercial Bluetooth detectors on a freeway under real traffic conditions. The resulting BTMS provided highly reliable TT estimations with a 5-min resolution.