I don't really see a contradiction here. The idea is that perfectly rational agents would choose to merge into a singleton that maximizes a combination of their utility functions, instead of wasting resources on competition. CEV is not the only mechanism that can achieve that, also see Carl's paper on superorganisms. Humans can't quite do that yet because we don't have good technologies for cooperation, precommitment or self-modification, and also because we don't have good enough math on bargaining which is necessary for merging.