It was raining yesterday so I got to visualize my sedan's rear window airflow. The sides dried out from the airflow while the middle stayed wet. All the way to the top of the window. I sketched the part that stayed wet.
I think this supports the idea of air vortices swirling in from the sides. With the sedan's less-attached flow coming down from the roof, the side vortices have more opportunity to swirl. (I think)
I think this means you're on the right track to consider the horizontal taper as well as vertical. I'm visualizing something more like Phil's actual boat on the truck - tapering in from the sides and down from the top, so it comes to (or toward) a point at the rear. The sides of the trunklid would remain exposed.