OldXCGuy, you seem to have the passion for race course accuracy that drives many of us that measure courses for certification/record validation purposes. It's quite frustrating to see someone run a 'record' or 'personal best' on a course that later turns out to be quite short, whether due to incorrect measurement, mismarked arrows, mismarked start and/or finish line, moved or missing safety cones, faulty navigation by a lead cyclist/car, etc.
If you have decent bike riding skills and the experience in running courses that's needed to identify the shortest path during a measurement ride, you can and should become a measurer also! This will also help you to scientifically explore your curiousity about how accurate your GPS devices are (you may be interested to know that two different GPS devices, worn by 2 different runners, will likely get different measurements for course length as well). Heck, if GPS manufacturers were really touting accuracy, they'd have a mode that would show distance to the .00001 of a mile, right? Forget that .01 crap, that's 52.8 feet!
You need a device AND a measuring method that is sensitive to small differences - think inches. That's what we use. If our two separate measurement rides are too far apart, we have a problem. However, remember that we start with a system the reliably measures distances to about 4 inches per 1/4 mile (and some use a system that measures to about 1 inch per 1/4 mile), so if our rides measured too far apart, we have confidence that there is a problem somewhere, and we have a way to find it and correct it.
Those of us who measure courses for accuracy want overly short/long courses identified and corrected. You make a case for having found one in NJ - take the next step. Get the map of the course, and compare the route traveled by the runners to the route specified on the map. Was the finish that day at the wrong location? (for example, the map may state that the finish is 112 feet, 4 inches north of the northern edge of a sewer grate on the east side of the road, but someone measured 112 feet, 4 inches south from the sewer grate on the west side of the road because they hadn't had their coffee yet) Was the 3 mile mark at the correct location? Did someone misinterpret a paint mark from something unrelated to the race as the finish line paint mark? Even if a course is certified, mistakes can happen. We'd like to find ways to avoid those mistakes, too. This is not a 6-Sigma process.
If the first 2 miles agreed with your GPS, then the SCPF would not be at issue, since it applies to the ENTIRE course measurement, not just the end. For example, the mile mark should have been about 5.28 feet after what your GPS recorded, the 2 mile mark should have been about 10.56 feet after what your GPS recorded, and the 3 mile mark should have been about 15.84 feet after what your GPS recorded. I suspect the finish was just mismarked.
In the early part of the pose, you seem to be confusing precision with absolute precision. Just because two different bike rides don't agree to the inch doesn't mean that a course is not accurate enough for certification/record validation purposes, or that the method and technology used is not precise enough for same.
For example, when I measure a 10K course, the 2 rides are typically within just a few feet of each other. That means, over the course of the 6+ miles, the measurements were within 1-2 walking steps! Some of the 11 items I mentioned may be involved, or others may be involved. Or, when I recalibrate the bike after the measurement, I may find that the two are actually closer than I thought because of the temperature differences in the bike tire. Usually, all I have to do is determine which of the 2 measurement rides is shorter and, if needed, extend the course a few feet with a steel tape. I've even measured courses where the 2 rides were within 8 inches of each other!
Again, you read like you have the passion and experience needed to measure courses for certification purposes. If you have precise bike handling skills as well, I hope you join us!