These are pretty interesting figures. If I've understood what you wrote correctly your conclusion would be that the screens that we love to hate really do work. Do you have any thoughts on situations that are particularly suited to such an ultimatum approach?
I don't think I'm enough of an expert in the technique to give that kind of guidance.
If somebody that was paying me asked that question, I'd try to figure out which of these "social cues" would work best for a sites demographic/customer base.
Then, define test cases. Think them through. Make them meaningful because if your site isn't VERY popular, you have to run these tests for a while to gather statistically meaningful data.
Then, TEST the heck out of these implementations. Use a service like BrowserShots. Make sure that whatever social mixins you've created work flawlessly. This will take more hardening than I'd usually do, past what I'd normally call "production ready." Remember, if there's a bug here, your site is just down for that user.
Finally, just test, test, test. And only trust the data.
