By the way, the small v's going further is probably caused at least partly by the way minecarts simulate friction. The faster a minecart is going, the more energy it looses. Therefore, sending the minecart down the long slope gives it lots of momentum causing it to loose a lot to friction. I'm going to do a test with the minecart going up first and see how that affects the result.
Edit: On the optimal powered rail configuration, did you perhaps change the direction in which your track was running between videos? As seen in the collision testing, that obviously makes some difference when dealing with minecarts.
Edit2: I think that the bottom cart went fastest on the up because it had the most boosters on the top flat section of any that made it up the hill (just a thought).