Without more data on that trail, hard to say. Three miles is short, though, compared to the examples noted by others however and the controlled experiments done by scientists.

Halfmile collected his own data on the PCT with very high accuracy equipment. (He did that sort of thing for his real job)

Makes a big difference then. The experience and knowledge to go with the high-end equipment too as well. As an aside, most of the mileage issues come from people doing their own tracking and questioning established routes or races. Not so much questioning of apps and its data by people who know what they are doing.

Halfmile collected his own data on the PCT with very high accuracy equipment. (He did that sort of thing for his real job). Guthook/Atlas used Halfmile’s data. You’ll notice that they have the exact same measurements to the 100th of a mile for their landmarks. Halfmile just updated the maps for 2018 including the added few miles in Nor Cal due to the new and improved small section that was done last year. Not sure when Guthook/Atlas will make that change, they might be slightly different for a bit north of there if the updates aren’t made at the same time. -GoalTech

The app and the mileage data they use are going to be two different things I believe. Atlas Guides (Guthook) uses Jerry Brown’s (Bear Creek Survey) GIS data for the CT and CDT apps from what I understand. Not as sure about Half Mile where he gets his data from as the source.

CopperTone

Your experiences don’t match what is being found in the field..or a lawsuit filed against a race director. :O
EDIT: And when professional surveyors measure distance for something non-legally binding, they use the mapping grade GPS..which are rather portable.

Recreational GPS units measure distances finely – inappropriately finely given the amount of noise in the signal. You can actually get quite good results – certainly adequate for putting the distances in a trail guidebook – from a smartphone GPS, but you need some sophistication to do it. If your track log doesn’t include all the data about signal quality (received signal strength, circular error probable, horizontal and vertical dilution of position…) it will be much more of a challenge to clean it up.

The trick is to take the track log, smooth it, and then interpolate it on the scale of a surveyor’s chain or measuring wheel, because that’s the “gold standard” for distance traveled, whatever a more precise instrument measures.

I did http://www.nptrail.org/?page_id=59 that way, and it matches the guidebook distances pretty darned closely except where the trail has been relocated or the guidebook has obvious mistakes. (There was one place where the guidebook simply appears to have got the figure wrong and come out exactly a mile short.) I don’t trust the last digit of the mileage one bit. I included it for calculation because I was accumulating roundoff errors otherwise.

And it’s sure easier to have one guy walk a trail with a smartphone and external battery than to have a four-person crew with alidade, planetable, sighting rod, and chain or wheel.