maximization problems are characterized by searching over a set of potential solutions and generating a proof that a particular solution is the best of those solutions

You're just saying that in this framework the function-to-be-optimized does not contain search/optimization costs. I think for most real-life optimization problems it's a shortcoming :-)