We report a good agreement between the shapes of tailored pulses obtained theoretically and experimentally by using the optimal-control theory and the closed-loop learning technique to maximize the ionization yield in NaK. The theoretical pulse shapes are found to be robust regarding the choice of the initial guess. We assign the leading features of the pulse shapes to processes underlying the optimal control and reveal the mechanism which involves an electronic transition followed by a direct two-photon process and sequential one-photon processes at later times. We show that the optimal control not only serves for maximizing the desired yield but also as a tool for the analysis and the identification of the responsible processes.