On the consistency of seismological models of the core-mantle boundary