I don't think number of games started in the last hour is really indicative of popularity in the first place. Number of games started * average number of players per game * average game length would be a much better approximation of a given map's popularity. The reason Battlecraft is at the top is because most games last 2 minutes and only 4 people can play, so new games get started every second. Meanwhile, hour long 12+ player maps like Risk get stuck at the bottom.
A somewhat randomized popularity sort would be good; give more popular maps more chances to be ranked high, but don't guarantee it. Something with popularity of 1024 could be only 2x more likely to be ranked higher than a 512 popularity map, and only 6x more likely to be ranked higher than a 16 popularity map. Since there are more unpopular maps, they'd fill most of your slots.