Puddleglum, I would also like someone to comment on this, since I like to have more than one source of info.
I checked once more to make sure, and that chart is for drag (Cx), not lift (Cz). As mentioned, for sedan and sloped-back (notchback and fastback) cars, the sloped C pillars cause a pair of eddies (I posted a scan of them somewhere, but can't find it now), while the diffuser produces counterrotating eddies that cancel. The angle of the diffuser determines how strong the countereddies are, and at about 3°-4° is the minimum, after which the diffuser's countereddies are strong enough to survive cancellation and bring drag down again. In a squareback car there is no sloped C pillar to produce eddies, so the diffuser's eddies have nothing to cancel with, and their drag increases more of less proportionally to the angle.
I'll update this later today when I scan another drawing.
I the mean time, here's a related thread:
Underbody panel testing
EDIT: Here are the promised drawings:
On the left is a sloped-back or notch-back vehicle (notice cancellation of eddies and countereddies), on the right a square-back vehicle (no countereddies to cancel).
Wakes (and eddies) from different vehicle shapes.
Drag (Cx) and lift (Cz) change with diffuser angle. This is from a chapter on race and performance cars, so I don't know how close it is for non-race vehicles. Based on:
Bearman P.W., De Beer D., Hamidy E., Harvey J.K.,
The effect of a moving Floor on Wind-Tunnel Simulation of Road Vehicles, SAE 880245.
Drag (Cx) and lift (Cz) of different downforce devices:
1 - Spoiler,
2 - Rear diffuser,
3 - Diffuser plus front air dam.