Self-voting at post creation time in HF19 gives 100% to the author, while reducing the curation share for every other voter. Even if a self-vote is done after the reverse auction time when 25% of the vote value goes to the curators, there are still 75% of it going to the author. Arguing with "sweetening the curator pot" is questionable even in this case, but that's a topic on it's own ;)
You were talking about a potential 1-2.5% increase in curator rewards between HF 19 and HF20 depending on when everyone voted. I was thinking that if that 6.6% that might go back to the rewards pool in your example actually ended up with the curators, it would be better to get the 25% distributed among them than just the 18-20%. Sweetening the pot might not be the right phrase, but potentially having the entire 25% to work with rather than only 18-something percent would be better than nothing.
As it is, I'm not sure why a 65-35, or 60-40 split hasn't been proposed, discussed, implemented, what have you. I understand the 50-50 split that was tried went away, not sure why, but guessing author's didn't like sharing so much with the curators. Which is fine. I would hope that the author is putting in more time to create a post than a curator is spending to read the post and then decide whether or not to upvote it. Or for that matter, simply auto voting, so I'm in agreement that the majority of the rewards should go with the author. There's still a range between 75-25 and 50-50 that could be explored, but as you say, that's a whole different topic.
If the reverse auction part of the curation share would remain with the curators, we'd have a fixed 75/25% distribution of the post rewards. This would actually be possible with changing a single line of code. Seems like this is not wanted by those in charge...
So, this begs the question, "Why?"