The template or something very close to it is the best guide to reducing aerodynamic drag, this has been shown to be the case both in nature through years of evolution for both air and water and also in man made designs.
The two images I posted before are what triggered my understanding when I was playing with the NACA 0039 air foil equation:
The hemisphere shows perfectly what the template is based on a 3d transition from all sides, but the square shows something completely different using the same equation, and this is the part that most people are missing I think when using scaling.
The question is what is the ground reference in the scaling, or more appropriately what is the focal point of the design?
For the pure template as a 3d full form object in free air it is the centre line from tip to tail, as a half body in close ground proximity it is the centreline on the ground. With the square as shown when the template type equation is applied to all points, the corners lag because they have the greatest distance from the focal point, hence the longest tail profile, whereas the sides close to the ground are nearest to the focal point, hence shortest tail profile and come in quickest forming the odd shape at the rear.
If you raise vehicle off the ground somewhat then there is a question as to whether the centreline on base of the vehicle then becomes the focal point and this would produce a shape somewhat like the Pinnafarina "Banana car".
As for apendages and attachments, out hanging free mirrors are basically seperate objects and can be scaled according to their size, scoups, shrouds, mudguards etc attached to flatish surfaces on vehicle can be scaled as if the vehicle panel is ground and this is their focal point.
As for a full boattail on a conventional vehicle, the template provides a guide for the profile, the sides theoretically could come in much quicker, but only if the corners are apropriately scaled as well, and they would present the longest tail.
This I think is the cause of the counter rotational vorticies, not the sudden meeting of top and side air, but simply the fact that the corner lines are curved in too quickly and there is not enough time for that air to decelerate in an orderly fashion.
Using the profile taper for the sides as well as top and gentle rounding can minimise this as there is an excess of air on the sides then and some of this can flow over and fill the void being left on the corners.
One of the ideas I have been thinking about is whether this problem can be solved by using a volume calculation using crossectional area and applying the % change in angle as set out by the template to the change in crosssectional area, so effectively merging the squareback at rear of vehicle to eliptical at end of boattail.