Different cost for same route?!

[jot ess]

Hello,

I'm experimenting to find the (physically) easiest bicycle route to my destination. My understanding was that this is more or less brouter's default. 

And I assume that "cost" is what a cyclist actually feels in his legs, so minimizing cost under the given profile's constraints is what brouter does.

I use the profile fastbike (in German: Rennrad, I guess this is the same).

With "fastbike original" I get a suggestion with 24.2 km, +171 m, cost 31378.

When I choose "fastbike alternative 1" I get a different suggestion: 21.6 km, +213 m, cost 37037.

Now comes what I don't understand:

I switch back to "original" (24.2 km ...) and add a waypoint as a constraint that makes brouter choose (almost*) exactly the same route that it suggested as alternative no. 1 and it results in: 21.6 km, +212 m, cost 32.982. This is a difference in cost of 4055! 

*(The differences between the routes are really marginal, like in the city in flat terrain first left and then right or vice versa.)

I have checked this with different examples, like fastbike alternative 2 and 3 and then constrain fastbike original via waypoints to choose these same routes. This way they "become cheaper" than they were as unmodified alternatives 2 resp. 3. They are still "more expensive" than the route chosen by "original", so their relative order is still sensible. But the absolute numbers leave me clueless.

If I find this unlogical, this is probably because I haven't understood the concept of "cost" the way this term is used in brouter. Could anybody enlighten me what cost actually means and/or where the above effect comes from?

Best regards

Joachim

[jot ess]

Another strange thing:

The above was my result with brouter-web. Now I had Osmand on my phone with the Android version of brouter do the same and a route that had +280 m of altitude gain in brouter-web has +370 m in Osmand. Might Osmand use its own (very much differing) altitude information to analyse what is has been served by brouter?