They may not use your elevation data exactly as it's recorded in your device log. There is probably a lot of massaging of the data to eliminate what each site might consider errors or noise.

Just pick a site and use the data you think is more accurate. Over the course of many rides, it might average out to being close with either site. To take one ride and compare it on multiple sites is just as crazy to me as assuming your device is even correct about the elevation gain and loss you had.

Accuracy really isn't as important as just knowing one route is more climbing than another. And you can figure out that by adding and averaging out the rides of the same route. Throw away what looks like outlier data for any one ride.