To clarify, because apparently that's necessary, how vigorously somebody starts/stops, or navigates their turns, depends on a myriad of other conditions that the GPS is simply incapable of recording. Road temperature, ice/snow/rain, other drivers, pedestrians with a desire to tempt fate, wildlife running out into traffic, etc.. The fact that you stop suddenly is not an indication of your inability to drive safely, and could actually be an indication that you're a *good* driver (a bad driver might not react at all and end up hitting the dog that ran out into the road). To use a GPS in that way is bad science at best, and negligent at worst, because it has the potential to reward bad driving habits (such as driving too slow or not paying proper attention to the road) while penalising good habits (such as reacting to sudden changes in driving conditions).
I'm still unsure as to your point?
If the data they collect shows that for the majority of drivers this indicates you're more likely to be involved in an accident then you're placed in an appropriate group by the insurance company for the risk they can assess.
A good driver may occasionally be confronted with a car that cuts them off or a pedestrian that forces them to break sharply. However a bad driver would statistically face a similar number of such obstacles while having more hard brakes due to driving too fast and too close. The good driver will still appear statistically safer to the insurance company.
I don't think the insurance company is claiming it's perfect. But I don't think your argument that some data may be imperfect means none of the data can be used stands muster.
The insurance company knows random stuff happens, but over 12,000+ miles per year, random stuff should average out. If you're frequently slamming on the brakes because of pedestrians running in front of your car, then however good a driver you are your risk of an accident is higher than mine and I see no reason why your premium should not reflect that.