I believe that 50/50 rewards would actually increase rewards for quality authors (bring back proof of brain?). I see the rise of bid-bots as a natural response to an imbalance between author/curator rewards. In effect, because returns to curation aren't high enough, bid-bots arose as a way for curators (SP holders) to 'capture' author rewards and put them in their own pockets.
You ran through the math on this very elegantly, so I see no need to add further calculations. I can mention that APR on leased delegation clings very closely to 100% self-vote APR, which provides support to the argument.
Generally speaking, I do not support protocol-level changes unless they cannot be done at condenser-level. Since a condenser could already raise curation rewards to 50/50 using beneficiary rewards, this becomes a grey zone. Would voluntary shifts to 50/50 be attractive enough to authors if they're not enforced on everyone at protocol level? I'd still like to see more experimentation with this before I would be 100% supportive of a protocol level change, but I recognize the opt-in nature of such experiments would limit their efficacy.
I would like a trial of some sort run that can be reverted. A few months isn't the end of the world.