Area measurement validation: ellipsoid vs geoid

I try to validate area measurements on orthomosaics using objects of known size on the ground. I’m working on centimeter level measurements.
As far as i understand an orthomosaic in WGS 84 is projected to the reference ellipsoid surface.

I set up a square of 1 m² on the ground which is level. Because it is level it means that it is parallel to the geoid (or more precicely to the local surface normal vector of the geoid).

By defintion ellipsoids are only an approximation of the geoid, which means the ellipsoid plane is not necessarily parallel to the water-level. This in turn would result in a error in area measurement.

Because RTK GPS systems usually use a bubble level for setup i would assume that this error is small. But how small is it? Are there locations on earth were this is a problem?