My paper has been provisionally accepted, but my latest results are not as reported in the paper. What should I do?
A good review panel would not select your paper purely due to the impact of your result but also with respect to your approach to solving the problem. This may include how existing methods were modified and how novel methods are applied. This would also concern your analysis of the methodology along with related results.
Withdrawing the paper (option 1) should be the last thing you ought to be thinking about. That is unless you plan to publish it in a different journal after much improvement, but this also has its own implications.
There is nothing wrong with going for option 2. In fact it is to be most welcome in the ethics of research. Hopefully the reviewers would also appreciate such honesty.
Most may prefer option 3 as this would mean that your paper is published without any further ado. But since this makes you -- the author -- uncomfortable, you rather don't do it. It is not worth the feeling of regret once it gets published.
Convey your points precisely and see if you could try to describe the fallacy of the overly optimistic result in your communications. You may also include this as part of the updated manuscript if possible. Note that all communications to the reviewers must pass through the editor as part of the procedure.