An operator-theoretic approach to the mixed-sensitivity minimization problem