I don't think you're missing something, although there are many nuances here. Selection certainly converges on local optima, but yes, there's nothing making a "choice" to do so (except with sexually selected traits and mate choice). It's just how it works.
In general it's unsurprising that selection misses hypothetical universal optima for the same reason it's unsurprising we didn't evolve wings even though that'd be an optimal way to escape lions. Evolution moves traits around fitness landscapes heavily contrained by exaptation. But interestingly many highly conserved mechanisms, e.g. essential cell machinery or metabolic pathways shared widely across the tree of life, have been shown to be maintained at absolute energetic and informational optima.
In general it's unsurprising that selection misses hypothetical universal optima for the same reason it's unsurprising we didn't evolve wings even though that'd be an optimal way to escape lions. Evolution moves traits around fitness landscapes heavily contrained by exaptation. But interestingly many highly conserved mechanisms, e.g. essential cell machinery or metabolic pathways shared widely across the tree of life, have been shown to be maintained at absolute energetic and informational optima.