Help!: Elevation profile changes with route length

I have noticed that I get very different elevation profiles for the same stretch of road depending on whether that stretch of road is part of a long or short route.  If I create a 20 mile route in a hilly area, the data looks like what I would expect.  However, if I create a 200 mi. route that includes the same 20 mi. section, all the smaller ups and downs disappear and only the gross profile remains.  I realize that the elevation data is limited to 500 points.  Thus, I would expect some decreased resolution with the 200 mi. route, since elevation data is only being sampled once every 0.4 mi. instead of once every 0.04 mi.  However, it appears that in addition to being thinned, the elevation data in the 200 mi. route is also being smoothed. For example, on my 20 mi. route, there is a section of road, represented by about 37 points, that climbs 200 ft. in 0.6 mi. and then descends 100 ft. over the next 0.9 mi. This same stretch of road in the 200 mi. version is represented by a single 1.6 mi. climb of about 100 ft. The end result is two different routes with completely different elevation data for the exact same locations.

I can certainly see the advantage of smoothing elevation data to eliminate noise.  However, the mileage interval over which you do this smoothing should not be dependent on route length.  I would expect the same stretch of road to produce roughly the same profile, regardless of the length of the route that road happens to be part of.  I wouldn't expect a 1.5 mile hill to be completely eliminated because my total route is 200 miles instead of 20 (if anything, I am MORE likely to want to know about a 1.5 mile hill after 150 miles of riding than I am after only 10 miles). If it is necessary to represent a profile with less points for a longer route, then just thin the points.  However, I think the best solution would be to report elevation data using a set distance between points and then increase the number of points to accommodate longer routes.