Yeah, consumer GPS resolution on an instantaneous sample could be off by 20m vertically. On a phone, if the hills are trending down for each climb in a set of repeats, it's likely an effect of the GPS constellation's orbital changes either decreasing or increasing accuracy during the set of repeats.
On a Garmin, this would be due to barometric changes during the ride (a front rolling in, for instance).





It's my job to understand the basics of GIS for various reasons


