What humans consider optimal for agriculture and what is optimal in the wild can be different. And evolution is not a perfect optimizer, sometimes it has to maintain legacy structures because what it cannot do is huge refactorings, only incremental changes.

The article also mentions RuBisCO being partially synthesized from chloroplast DNA, which limits evolutionary speed.

Nature does evolve improvements (e.g. better rubisco in microorganisms, the C4 pathway in some higher plants) but other plants can't just download the updates from a central DNA package library.