Explained much better than I did!

Yeah, Henry nailed it.  

The one aspect of this that's worth highlighting is tunnels.  The surface maps that record altitude for specific GPS coordinates are, well, "surface" maps.  So when a road goes through a tunnel, the altitude that gets used is the altitude of what's on top of the tunnel.  Not the altitude of the road surface.

This probably isn't news to most RideWithGPS users.  We've probably all wondered why there's a bump with a 35% grade in the middle of our routes.  

Unfortunately, these bumps get calculated into the climbing statistics for the route.  Another reason to treat those statistics with a big grain of salt.

I find it useful to turn on the "Grade" option at the bottom of the RideWithGPS route planner.  It's a great way to find potential routing problems.  In the case of a really high grade (for us, over 12%), you know you've got something that needs more investigation.

Sorry if this is too well-known - just hoping it helps someone down the line.

Kelly my most recent tour (Canberra to Hobart) was my first attempt to seriously use a Garmin GPS, with downloaded routes from RWGPS and Kamoot. I found I also needed to frequently cross check with Google Maps. Variation between the three sources was significant. 

The old saying is that “the map is not the territory” and it quickly became clear to me that those navigation app maps need to be treated as a helpful tool, but not relied on. 

For what it’s worth, Google seemed to be the most accurate for elevation, but only if I could convince it to choose the route I’d prefer to ride. Sometimes this meant juggling Google Maps between the walk, cycle and drive route option to get an estimate of total ascent. 

My cyclocomputer's pressure-based altimeter USUALLY measures slightly less ascent than what RWGPS indicates for the same route. My actual ascent closely matches RWGPS in open flat or gently rolling terrain.

RWGPS elevation gain gets inflated when roads go alongside cliffs and include bridges or tunnels. Bridges and tunnels seem to be accounted for properly on major highways, probably by a manual correction process. But back road routes routinely show unreal giant elevation changes at bridges and tunnels.

Comparing predicted vs actual elevation gain of the tour I just finished, most days my actual elevation gain was 1% less than RWGPS predicted. But a few days had actual elevation gain 20% less than the RWGPS prediction. Like I wrote previously, routes alongside steep cliffs seem to have the biggest discrepancy.

I am no expert in mapping, but I suspect that a .GPX file contains 2-dimensional route information and little or no elevation information, and that each site calculates the elevation changes of a route using its own preferred geographic database.

Hi Graham 

I am usually in full agreement with you but I have to take a contrary view of your statement the "Google seemed to be the most accurate for elevation".  Our experience,  in various parts of the world, has been that it tends to under-calculate elevation and seems determined to simply state "Mostly flat" whenever possible. 

Today's ride from Chilecito to Miranda in northern Argentina is a case in point.

To start off with, let me show Komoot's data.  The climb from Nonogasta, the low point for day, to our digs in Miranda is a constant climb with a net elevation gain of 540m and a total climb of 550m.  A very constant climb with few dips.

Google's cycle route between the same two points selects a slight detour onto a no-longer-existing road, records no elevation gain and states "Mostly flat" despite the start and end points being identical and an undeniable climb of over 500 meters. 

Manually correcting Google's route so that it sticks to Ruta Cuarenta (the iconic 5000 plus kilometer road stretching the length of Argentina) causes Google to now recognize the climb and stating "Moderate hill" with an elevation gain of 542m and a descent of 2m.

Jean-Marc thanks for such a comprehensively researched reply. You’ve gathered a lot more experience than me. As I mentioned, my last tour was my first serious attempt to use a GPS unit properly.
I need to practice more.

Another app I did refer to occasionally was EarthMate. It shows topo lines which can be useful.