Hmmmm... This seems like something that could work. But it reminds me of what Sirrian stated before:
Let me explain what i understood and you disagree you can get a Crazy Ticket with @HKdirewolf and an Idiot Ticket with @htismaqe so you can get in line to me call me that.
I'll show a sketch of three bracket with five ranks each in your model and compare the rewards with the total amount of the rewards from Sirrian's table using only the gems as an example:
First of all, i didn't tried more than once to fit the same amount of gems corresponding to Sirrian's table because i wanted to do it quickly (And the numbers may be tweaked anyway), but i did matched the first rank of each bracket to the same amount of rewards for the last rank of the next superior bracket.
Why you did that?
Well, it wouldn't make much sense if your guild arise into a new bracket, even to the lowest rank of it, and still get LESS rewards than being at the top rank of the previous/inferior bracket. This is also sensible to guilds that would move out from a superior bracket into the highest rank of the next inferior bracket giving some chance to recover the position.
Even if the team tried to give some epic rewards to a few guilds i believe the players in general are more inclined towards evening the distribution. And personally i don't mind about the rewards, i liked this concept, as i understood and how i worked on some small details in this sketch, because it gives mobility, as guilds could recover from a given week where they underperformed as well as it gives visibility as some guilds overperforming start to climb the ranks week by week. Sounds good enough to not become an stale/boring feature.
This is how i'm assuming things should work on your model and tossing some numbers to make it more visible. And hopefully it can be implemented.
You don't need my approval or anything, but i was pointing out a flaw in your suggestion and that's all. For the highlighted part i believe that we must try to do better and never settle for the same mistakes in the past.