This is actually not true. Let's say a police officer measures 100 cars going at exactly the speed limit and the measuring device had a 1% chance of showing a speed 10mph above the speed limit. In this case, the officer will end up ticketing a guy who was actually going AT the speed limit with an inaccurate reading. 100% of the drivers who are ticketed in this case are victims of the reading.
This is the same effect plaguing psychology. Over half of psychology papers are not reproducible even though their results are accurate at 95% confidence intervals.
http://www.nature.com/news/over-half-of-psychology-studies-fail-reproducibility-test-1.18248
I have yet to have someone assist with providing me this information. I did find a consumer grade device with some info. Assuming a normal distribution (the best I have to work with since they don't tell distribution!), there is a .03% chance on each reading that the measurement will be 12+ MPH greater than the actual speed of the car. To put it in other words, if I took 300 readings, most likely I will measure one guy going 12MPH faster than he actually was.