They, as a guild of real life players, get to "win" more matchups/seasons.
Let's say you have a single guild alliance of 500 players, that they average 10 hours a week per player, and they all have 2 accounts.
And for ease of setup, let's assume anet's averaging period is the same as the matchup period. The matchup period and averaging period can be different, but all that does is smear out the inequalities in team formation, it doesn't "fix" it. We'll also assume that this guild has been hopping around looking for fights previous to alliances, but will try to "win" once alliances start.
First alliance matchup, the 2 guilds are assigned 2 worlds, each world is charged roughly half, 2500 player hours, to have these 2 guilds as part of them.
The alliance picks one world, and plays only that world to "win". That world now has +2500 player hours, and whichever world they don't play on, has -2500 player hours. Since the "winning" world is getting the full 5000, and the other is getting 0.
Next 8 week period, a new set of worlds is created. The guild this alliance just played on charges 5000 player hours to the world it is assigned to, the guild they didn't play on is charged 0. At this point 2 scenarios can occur:
1. They play the same guild again, anet's system "worked"
2. They swap guilds, anet's system just utterly failed, since now there is a world with 5000 extra player hours, and a world missing 5000 player hours.
As you increase the number of guilds you can choose from, you benefit more, and decrease the odds that all of your guilds are assigned to the same world. Further, multiple alliances can coordinate to exploit even larger advantages, by coordinating to only play on worlds where multiple of their alliances are paired up.