First of all, thanks for such a good test case!

Elevation calculation is always harder than it looks. Although the natural thing to do is to read the elevation of every point along the route and add it up, the variation in the elevation grid data makes it seem like there are many more tiny ups and downs than there actually are – particularly in steep-sided valleys. So cycle.travel does a lot of smoothing to try and get a more realistic result.

In this case, the smoothing wasn’t working properly. I’ve reworked a few bits of the code in the light of this and it should now be reporting a more accurate result – it now reckons 5048m which seems much more likely.

For the Garmin issue, this is probably because the Garmin unit is just robotically adding up the elevation numbers, without doing any smoothing. If you go via Garmin Connect, I think that probably does some smoothing/correction before sending it to the unit, so should come up with a better result. I’m working on Garmin Connect integration at the moment, so give that a week or two and we might have some progress!

Thanks. +5000 is probably closer to reality. In fact, I know the route very well and know for sure that there is one single spot where the route goes down then up, adding some elevation above rough passes elevation calculation : probably not +500 compared to +4500.

Thanks anyway for digging into my case. And thanks in advance for the integration to Garmin. In find Cycle map much more userfriendly than garmin connect