There are a lot of ways to measure trip times, and thus average speed, and they vary a lot depending on the trip. There will always be debate over what is fair. For example, a car leaves when you want it to, a train may only leave every 30 minutes. Should we add an average 15 minutes of wait time, or figure that people just alter when they will leave based on this. (Once transit gets to about a 5 minute interval, people start to ignore schedules and just leave when they want.)

For those who aim for a specfic train, we must add time for when it is late. Any time you have a transfer, you must factor in not only the typical wait time but the occasional risk of a missed connection.

This (like traffic jams for cars) also affects the other thing people care about -- standard deviation. For example, if the train usually takes 30 minutes, but one time in 5 it takes 45 minutes, then if you have a hard deadline to hit you typically must plan for it to take the 45 minutes (and accept the consequences of lateness if it takes 50.)

This is a killer when trying to hit really hard deadlines like plane flights or infrequent long-haul trains. Unreliable transport to the airport forces you to allocate a huge time for a trip that might actually have a short average time.

Of course, dedicated right of way is great for making the time reliable, as well as improving the average time. I suspect the reliabilty of dedicated ROW is more important than the speed gain. This is one of the things that attracts people to PRT over cheaper and better established at-grade light rail, or even dedicated ROW light rail that still has to deal with traffic lights.

It's possible that the real solution will be the arrival of crash-avoiding, self-driving cars as PRT, using existing roads with some special privileges (such as private lanes.). But until those arrive, people will focus on other systems.