On a consistent mean spherical model for plasmas