On a relationship between distorted and spectral risk measures