But if I don't - if I'm inclined to treat data on overhead and estimates of utility as very noisy sources of data, out of skepticism or experience - is it irrational to prefer several baskets?

Very much so. Rational behavior is to maximize expected utility. When rational agents are risk-averse, they are risk-averse with respect to something that suffers from diminishing returns in utility, so that the possibility of negative surprises outweighs the possibility of positive surprises. "Time spent reading material from good sources" is a plausible example of something that has diminishing returns in utility so you want to spread it among baskets. Utility itself does not suffer from diminishing returns in utility. (Support to a charity might, but only if it's large relative to the charity. Or large relative to the things the charity might be doing to solve the problem it's trying to solve, I guess.)