GPS isn't always as accurate as you'd like it to be. Even DGPS, which is what a/c use, and surveyors use, needs time to accurately decide where it is and whats going on around it.
http://news.idg.no/cw/art.cfm?id=650277B6-1A64-67EA-E43C4F57008DA7A1
the system (of which, there really are three) is so old, it is going to begin to degrade.
RADAR - cosine error only exists when a target is not parallel to the beam. If the horn is parallel to the roadway, there isn't any cosine.
RADAR is accurate if used properly. A nick on a tuning fork isn't an issue as long as its' not a particularly deep one. Tuning forks are NEVER used to calibrate a RADAR unit, only to verify correct operation. End users NEVER calibrate a RADAR unit; they are (supposed) to be sent to a shop periodically.
Even so, case law doesn't support using RADAR in greater than a 1 MPH increment. In other words, if the posted speed limit is 55, it would be hard to prosecute (theoretically) a citation for 56 miles an hour.
Tire wear is a negligable issue. Part of your calibration verification is to match patrol speed with speedometer readout. Speedometers (should) be calibrated by a shop on a periodical basis, and the tires can NOT be different than the ones specified.
Lastly, until recently, no jurisdiction in the US allowed RADAR as the primary method of establishing speed. RADAR / VASCAR / LIDAR all are tools to assist the officer in confirming their inital visual estimation of speed. All NHTSA approved RADAR / LIDAR (dunno about VASCAR) classes include a field practical portion where you go out, see a moving vehicle, estimate the speed, then see what the RADAR says.
The best, most accurate way to estimate speed is by measuring a distance, then using a chronometer to time how long it takes to traverse that distance, then mathing it out to determine speed.
-Shawn