The biggest (from my point of view) problem with this request is that it can lead to abuse and becomes particularly problematic on meta.
Pinning has the side effect of getting the answer more attention. Even if the OP doesn't earn reputation by accepting the answer, they will still likely gain reputation because it's accepted and pinned and they're the one who controls that.
We encourage users to accept the best answer. Knowing that their own would get pinned and earn more votes might lead them to pick their own answer just for the visibility. It may be correct but not as clearly written as others, so assuming it would get downvotes isn't necessarily the case. Yes, this is always the case but with self-answers, the user has more incentive - the sweet, sweet 10 rep per answer upvote. As such, we require that the community recognizes that the answer is the best by voting it to that position.
Additionally, I've seen people write self-answers that are wrong or otherwise problematic. They generally get downvotes but they'd still be pinned. While it is possible to accept a bad answer written by someone else, it's at least less self-serving to accept it.
This is even more troublesome on Meta, where the content is more subjective. A user may have an idea for a feature or start a discussion and answer it. Accepting it makes it seem like the "correct" answer, which is confusing to people when they see that the answer scores lower than the others. Again, I recognize this is still an issue when the question is answered by someone else but that requires that someone else at least has that response. The pinning can be abused, even on child metas where there is no reputation.
For these reasons, I believe that self-answers should not be pinned.
To address your specific reasons:
- Accepted answers indicate the answer worked for the OP, no matter who posted it. Most people viewing a question want to see the answer that solved the problem best, which is what accepted answers are for.
While many people think that the accepted answer is the best answer, we don't actually define it that way. The accept indicates the OP's preferred answer. Nothing more or less. That doesn't necessarily make it better in any way, it's merely an indication of preference; that of the OP specifically. In fact, it's confusing to many users who try to accept several answers because they assume it's the same as an upvote. If it truly "solved the problem best" we wouldn't pin accepted answers at all and we'd instead go by votes only.
- It causes a lot of confusion and unnecessary bug reports. As mentioned above, there have been lots of bug reports about self accepted answers not rising to the top.
Yes, this does confuse people occasionally. I've been caught by it before myself. I'm not sure that confusion warrants this change. There are many confusing parts of the network. The solution is generally to do a better job explaining the situation than to change it how the system works. We could indicate why the question wasn't pinned to alleviate that.
- When the accepted answer isn't at the top, it appears that there is no accepted answer. It's really misleading and makes users think the asker hasn't solved their problem, and that the question needs more answers.
Many people get to the question from the search results page, which indicates questions with accepted answers by filling in the answer box with solid green, so they arrive already knowing the question has an accepted answer. Additionally, there's no correlation between questions with unaccepted answers and lack of solutions.
Many people who ask questions never return to the site to see that they've received an answer and see if it's correct. On the other hand, on some sites, there's a problem with questions having no answers at all and yet being answered in the comments. The two can actually go hand-in-hand. A question that's answered in the comments often leads to the asker disappearing and never coming back, even when real answers come in later on.
I will agree that accepted answers often deter new answers on a post but even having an accepted answer doesn't mean that a question can't or shouldn't receive new answers. Questions can often use more answers. Many users are hasty in accepting the first answer they receive, regardless of how good it is.
