My method for getting clean, highly accurate GPS tracks produces great results but is silly time consuming and utilizes ESRI ArcGIS, which is more costly ($100/a year for personal use, or $1000s for professional use if you're selling what you make like myself). I'll export all the trail tracks I have collected for an area from GAIA and Garmin, import them into ArcGIS Pro. For a given trail it may be a dozen plus or just one. Then I load up full resolution 1 meter bare earth lidar if it's available for the area, and whatever the highest resolution and most recent Aerial imagery is that I can acquire. I use my tracks as a guide, while also referencing strava heatmaps and OSM, and create an individual trail vector line in a database for each designated trail with vertices as often as needed to follow the trail as accurately as possible, referencing all those layers. Then add in user trails, spurs to viewpoints, etc. My least accurate areas are parts of deeper forest with no lidar coverage. The biggest variances I've found from OSM are of course in less traveled areas where the OSM data was derived from USFS data or tracing old topos, but I've made some much much cleaner routes of even popular trails like sections of the PCT or of course the Timberline Trail. Heres ahttps://elevationchanges.com/2022/01/08 ... ng-trails/ article I wrote up about using lidar to do this.
Having things so accurate really comes to fruition when I start playing with super zoomed in trail highlights, especially when brought to life in little 3d mini maps like I've developed for my Crater Lake NP map like this: