Mist trail "wind tunnel"
 

Mist often drops from the sky, thickly, this time of year here in Southern California (think of a classic Humphrey Bogart foggy scene... it was shot near where I live).

Take a look at photos 1, 2, and 3 in my photo album "Black and Green" and tell me what you think of the similarities and differences between the driver and passenger side air flow traces. Remember, I have no outside mirror on the passenger side: Fuel Economy, Hypermiling, EcoModding News and Forum - EcoModder.com - California98Civic's Album: black and green

Top speed was 60 mph, held for 10 or 15 seconds before coasting to this stop.

I think it suggests my mirror delete might not be doing much good. Maybe that it is even harming the aerodynamics of the vehicle. :(

I'm not sure what to make of it.

However I'm not going to assume that a clean removal of moisture off the rear window is the result of attached or non-buffeting air flow.

The passenger side may be more revealing because the vortex of air which forms as two different airflows and air pressures meeting at the rear pillars or blind-spots (over the top air travel and along the side air) is less disturbed because of side mirror delete.

Side-winds?

dry-wet-dry pattern
 

My hunch is that the roof is producing counter-rotating vortices at each side.
The mixing and shear of the highly energetic vortices is enough to evaporate the dew where they contact the body and backlight.
Down the center,where the vortices are helping to hold the boundary layer against the body,the straight-line flow does not contain enough energy to evaporate the moisture,leaving the pattern which you see.
Hucho depicts such a scenario in some of the schematic representations from his,and others research.
VGs would alter the situation.It's exactly what their special abilities are crafted for.

Jumping in there because my Saturn has similar markings after a light rain. I'm having a little trouble following your post, Aero, where would recommend the VGs be mounted on the car?

Zoom in on these images below to get an idea where you can delay the vortex forming and the associated drag which comes with them.

I'm not really sure that VG's can delay large vortex generation because they cause their own mini vortexes.

There must be an overall concept before application though, a method of testing ideas and proving them right or wrong.

VG location
 

I'm not really qualified to say,as I've never worked with them.
I went into the Phil Knox Aerodynamic Photos album,to the bottom library of articles/clippings and looked at the Wheeler VGs.It looks like they were positioned a couple inches ahead of the backlight surround.
If you can find the prior posts on VGs,and locate the Mitsubishi Lancer study published by Mitsubishi,it would give you some additional insight as to proper placement.
Technically,VGs need to be sized for the boundary layer thickness of the specific application,and it would require a wind tunnel and smoke/drag measurements,to actually dial in the sweet spot.Which is why I shy away from them.
If you had access to a really dusty dirt road,you might be able to use the dust patina to locate the flow separation line on the roof/C-pillars.
If so,you'd place the VGs just ahead of that line.
Lastly,if you compared the Aerodynamic Streamlining Template to your roof,the location where your roof just began to fall below the 'Template' contour would indicate the point where your boundary layer is beginning to be compromised,whereas the VGs could help to re-energize it at this location.
These are just educated guesses.:p

If your hunch is correct, then images 1 and 2 in the album I linked reveal that the passenger side--where the mirror delete is--has notably less turbulance or vortices. The driver side shows the drying pattern over the driver's window and back along the rear quarter-panel and further across the rear window and trunk lid toward the middle. Didn't you say once that very precise max speed testing on your fourth gen Civic showed no change in top speed between mirrors and no mirrors?

My car is clearly closest to "G" in that image you shared.

mirrors
 

Yes,on bracketed runs,the difference could not be measured to 1/1000th of a mph degree of accuracy.

So, even if the flow patterns had changed somewhat, with mirrors and without them, the total power lost to drag was not significantly reduced by the changes?