[RedDevil]

I'd like to comment on the sharp edge on the rear bumpers corners.
There is a reason why this is seldom done. The reason is that the sharp edge does not always benefit drag reduction compared to riounded corners. Also rounded corners allow the bumper to be stronger and lighter.

I own a 2011 Honda Insight. One of the complaints sometimes heard about that is its side wind behaviour. And I have experienced that too; but only in very specific conditions. Which ultimately led me to figure out (or so I believe) what is really going on.

First of all you have to know about the Coanda effect. Air, water, anything flowing in a steady motion along na edge will follow that edge if it bends away. If it didn't a vacuum would be created between the edge and the stream (and indeed, in the bend the pressure is lowest close to the edge; this is the force that bends the stream, or more precisely, the higher outside pressure does that)
The more stable the flow is, the stronger the Coanda effect will be. The Coanda effect destabilizes laminar flow; disturbances in the flow will thus be enlarged, as the parts of the disturbance that least follow the curve of the edge will have lower pressure than those that do, so they grow while the other parts of the stream dwindle.

When a laminar airflow along the side meets the rounded edge of the rear bumper it will curl around it, inducing a vacuum that pulls the car back and maybe even creating a horizontal vortex roll which low pressure center also pulls the car back.
A sharp edge makes the laminar flow continue in its original direction, creating a sharp divide between the air behind the car moving at more or less the speed of the car and the air just aside of that, which is pretty much stationary. The air behind the car would have pretty much the ambient pressure, so it won't pull the car back.

So why is the sharp edge so uncommon? Where's the snag?
The snag is that the airflow on the sides is usually not laminar at all. Straight against the wind, the same air that is close to the side at the front fender will still hug the rear fender. It will have met 2 wheel wells and creases at the door panels etc. It will be turbulent. It will pass the rounded edge without much interest; it is an unstable system to begin with.

With a very strong sidewind that all changes. The air close to the side body panels gets flattened like dough under a roller, and so are any disturbances in it. Now the airflow at the rear edge will be laminar and happily curve around the bumper. This will cause drag but it still does not explain the constantly jerking disturbances I feel when hit by sidewind.

The jerking behaviour stems from the sidewind being just not strong enough to produce constant laminar flow. The Coanda effect will kick in and break off violently, due to its instability enhancing nature.

The highways in Holland are designed not to follow straight lines for long but use very mild curves, sometimes deviating more than 45 degrees over several kilometers, to raise locational awareness and fight boredom. They also make for conditions to study a cars reaction to varying angles of wind at high speed.
What I noticed in my Insight doing 95 km/h (58 mph) was that with high wind (6 Bft) almost straight aside fuel consumption was at about 5 liter per 100 kilometer. While my direction slowly turned against the wind it suddenly increased to 5.5 to 6 l/km while shaking the car. Turning further against the wind, the shaking stopped and the FC went back to 5.0. Straight against the wind it even dropped to 4.5.

So the Insight is a nice example of a windtunnel optimized car that suffers in real world conditions. I believe the rear bumper is the main culprit here, especially its smooth rounded edges. Even just detaching it seems to help mpg.

What to do? I've got kids in the car so the bumper stays on. Reshaping it is a bridge too far for me.
Anything that breaks the Coanda effect should be good enough.
A small ripple (like created by a thin zigzag tape) could do the job.
I plan om sourcing colorless zigzag tape and stick it on just where the curve starts. If I can't get it then maybe a stripe of clear sticky tape encasing a nylon guitar string in the middle will do.
Of course there are many questions still to be answered about this, like G, B or E?
E obviously is the lightest but might not be enough to disrupt the flow.

My apologies for this lenghty reply, especially from an aerodynamical layman like me but I got this in my head and start to really believe it.
As I lack the skills to disprove my theory, I hope that someone who knows points out my errors and stops me from ruining my car and its mpg efficiency.