Fix HTML uses the HTML 5 algorithm to fix broken HTML. In other words it results in exactly the HTML you would get if you tried to display your broken HTML in a modern web browser. This is the only sane way to implement fix html. It means that the result of doing fix html will look no different than what the original HTML looked in an actual HTML renderer.
If you want to change it using search replace use:
search:
<p.*?>(Chapter\s+\d+.*?)</p>
replace:
<h2>\1</h2>
If all the p tags have the same class you can be more specific in the above expression