We originally added the first one to support the sphere-based Mercator projected coordinate system that is being used by several of the web-based mapping applications. The projection uses spherical math. The ESRI Mercator supports both spheres and spheroids/ellipsoids. If you have a PCS that is based on a ellipsoidal GCS, the ellipisoidal math is used. The results won't line up with the web service data.
So we created a custom sphere-based "WGS84" to use with our standard Mercator math. A sphere that uses the semimajor axis of an ellipsoid is a "Major auxiliary sphere." However, that causes problem because technically you need a geographic/datum transformation to convert to GCS_WGS_1984_Major_Auxiliary_Sphere--but the transformation shouldn't do anything. (Some transformation methods will change the latitude values between an ellipsoid and a sphere)
Later in the development cycle of version 9.3, we ended up adding a special version of Mercator: Mercator_Auxiliary_Sphere. It has a parameter (Option) which you can use to specify whether to use the semimajor or semiminor axis, or calculate an authalic radius, if based upon an ellipsoidal GCS. That way you can use the standard WGS84 GCS, thus getting rid of the transformation requirement.
Because it went in so late, the other teams couldn't switch to the 2nd version for 9.3.1. I believe you can only publish with the first one still.