I wanted to build a format where tri-color cards were not only a viable option, but also a natural way to draft. As Jason so eloquently puts it
here, supporting all ten tri-color combinations is asking for those cards to be picked last, simply because players won't be in the right colors. By reducing the available two-color combinations in my cube to five, and supporting only the corresponding shards/wedges, I ensure that it's much more likely that players end up in the right colors to support tri-color cards they see floating around. The likelihood of this happening is further increased by running more gold cards and a full 10% worth of mana fixing lands (including
borderposts).
Now, if you are not aiming to support tri-color cards, there might still be another reason to stick to five colors. I think it depends on how "gold" you want your environment to be. If you are planning on running a lot of gold cards, running all ten pairs is probably still a mistake, because you will end up with people getting "free" strong gold cards, simply because they're the only drafter in that guild. In other words, if you run all 10 guilds, you reduce the amount of "fighting" between drafters, and thus reduce the amount of tension during the draft (because you can reliably bank on certain cards wheeling).
Now, if you are planning on using gold cards (mostly) as signposts for an archetype, and plan on running only a select few two-color gold cards, I think running all ten guilds is perfectly feasible. You'll end up with only a few cards that will wheel to one drafter, but those will be the signpost cards that they are drafting around, so that's actually a good thing! Since most of the cube will consist of monocolored cards, players will be "fighting" plenty over those cards. As long as you make sure you don't secretly stuff your monocolored sections with cards that really only go in one of the signpost archetypes that is.