When I find those awful (IMHO) ePub with <div> instead of <p>, I replace the <div> with <p> where it should be <p>. And because the <div> and </div> are on separate lines, I search/replace all the lines so they run on. Then I use Modify ePub to split the lines properly. I will go through the code to see if I missed any <div> that should be <p>.
Does anyone know what software is being used to screw with <p>?