I don't know how your trackers worked, but I can tell you a couple of ways that have been tried with some success.
1. Triangulation from cell towers based on signal strength. The towers ID themselves so you know their location from mapping data, and in principle there are rings of constant signal strength around the towers. If you can hear three towers you can get an unambiguous solution for the location. Obviously antenna patterns aren't perfect and buildings and trees mess with signal, but generally these systems claimed accuracy to a few blocks or better. Probably more than good enough for this application, where you are just looking for folks skipping town. The five wire harness could have been carefully designed to provide cell phone reception.
2. Tying into the cars signal traffic for speed and possibly direction information, augmenting this with modest quality rate sensors and possibly accelerometers and then doing dead reckoning. There are a number of tricks to making it work. The cars info on zero speed is pretty accurate and those periods can be used to measure and cancel rate sensor drift and accelerometer offset. Built in maps can be used to test/correct the position estimate with the assumption that most motion is on roads. These systems claim to delivery accuracy measure in modest numbers of meters.