Mercator projection from the sphere was developed by Mercator, Wright and others, in the period between 1569 and 1640, when logarithms were fully discovered and understood.
When, years later, the ellipsoidal shape of the earth was proven, other scientists (Murdock 1741, Lambert 1772) developed the ellipsoidal formulas.
They are called direct Mercator formulas for the ellipsoid, but are not a Mercator's work, since Mercator died in 1594.
The same thing can be said for some other projection invented later, for example transverse mercator projection, and first of all, Web Mercator projection.
Web Mercator projection was created by Google in 2005 (the number is acronym 900913=GOOGLE), with the specific porpouse of projecting their maps in a simple way.
Since Web Mercator goal was Web visualization and not accurate surveying computations, the loss of conformality implied when using directly the simpler and faster spherical formulas (developed to project from the sphere) to project from the ellipsoid WGS84 instead, didn't worry his creators.
I can confirm that Web Mercator is not conformal, because I am doing a master degree thesis on the subject, and I've checked all formulas, including the scale factor one, that can be found in the EPSG's Geomatics Guidance note.
Being the scale factor at a given point, not a constant, but a function of azimuth (direction), the projection is not conformal by definition, technically speaking. But it is almost conformal! An infinitesimal circle drowed on the ellipsoid surface, would become an ellipse on the map, but an ellipse with a very small flattening, and really similiar to a circle. For this reason deformation of shape is very small too when looking at small areas, and is not visually notable; a square building is projected as a (almost) square building.
