Here is a decent and easy to understand article on GPS surveying:
Here is a CalTrans document on GPS aacuracy requirements for different surveys. 5 mm is the most stringent requirement and used for monuments, etc.
Makes me appreciate the building achievements of the Ancient Civilizations that much more; Assuming the Ancients weren't helped by advanced intergalactic visitors, or future Us traveling back in time to mess with ourselves in the present.
I thought that claim by the parks department person sounded a little hard to believe. I had previously heard that the best accuracy was about one centimeter but it looks like it's a bit better than that.
Food for thought:
After I saw the accuracy claim I called the surveyor I work with that surveys our deep water well locations (think surveying a spot 10,000' under water) subway tunnels, and surveys in the GPS systems on ships (ie he is one of the best surveyors in the world). He told me 2 things. The first is the most accurate survey possible is about 5 mm by waypoint averaging in 30 min increments over the course of the day to grab different satellite constellations and then performing post processing using tie in points for reference that have been surveyed multiple times over the years. The second is that there really is no "survey grade" GPS. Without the post processing our handhelds pulling off GLONASS, GPS, WAZ, etc are just as accurate if left to waypoint average for the same length of time. Without post processing he says the best accuracy to expect is ~ 3 ft. I confirmed this as he surveyed our farm for us with me clearing survey lines. At the GPS points he waypoint averaged at for 30 minutes and then post processed I can match within 3 to 6 ft if I waypoint average for 3 minutes.