I think it comes down to the fact that a hub and spoke model, where relative values only have to be maintained and agreed upon into and out of a single commodit

I think it comes down to the fact that a hub and spoke model, where relative values only have to be maintained and agreed upon into and out of a single commodity are easier for humans to conceptualize and agree upon at-large than complete (or worse yet and more likely, incomplete) graphs.