The GPS systems onboard drones (or cars) have, per se, not less than 5 meters horizontal accuracy (compared with terrestrial coordinales) and much more than that on vertical axis. Hardware innacuracies, atmospheric changes, ionic space storms, solar activity (Kp number you can check in NASA pages or UAV App, for example), electromagnetic interferences, etc. make GPS accuracy that low.
If we use much better (and heavier) equipment, GNSS systems are able to diminish such large errors by using ground stations, and you can obtain accuracies down to several milimeters horizontally and may be a couple of centimetres vertically. This is what we measure in each GCP (ground control point) qith a dedicated terrestrial system. A RTK system aboard a copter drone or fixed wing will give you few centimeters accuracy, but an accuracy of the position of the flying camera in that moment, not on ground.
When you input photos taken with a Phantoms or Inspire drones in Professional software like Pix4D, once the soft accepts the series of photos, they will show an accuracy of 5 meters horizontally (and 10 meters vertically) in the geographical coordinates which position the camera at the moment of each photo. This is what the expert professionals designing the software defined as standard errors coming from standard mobile GPS
Thus, achieved accuracy (precision) of flights made by drones over a terrain can be as precise as it they were made by hand and foot surveying the terrain. The real, and to me most important, difference is that traditional surveying method will give you one hundred (or several hundreds) very accurate points of a field, while aerial survey (plus GCPs) will give millions of points (the point cloud) all of them with the same accuracy, and each one of them with the color and luminance of the real object. Point clouds can, then, be seen as a 3D photograph, making measuring and identifying tasks much, much easy to achieve.