no, proposed solution is better. Giving the curation share from a big self vote to curators of that particular post is very easily gamed, and would be gamed. All an author has to do is set a second account to upvote the post at optimal curation time to get the reward from the self vote. Returning it to the pool is much fairer as it is impossible to game by the individual self voter.
RE: This has been a demonstration of optimal posting strategy under Hardfork 20.