They used a Kodak FF CCD and Nikon and Zeiss lenses to do the testing.
Vignetting occurs because the lenses can't provide full illumination at
wide apertures (the edge light rays are blocked by the internal lens
aperture).  This is allowed because of the aberrations that would occur
if that outer rays were allowed to pass.
Only way to correct this is with lens elements of larger diameter =
more cost.
The second problem are the microlenses used on CMOS and interline CCDs
because of their
poor fill-factor.  The light sensitive area of the pixel is much
smaller than the pixel itself.
Off axis rays hitting the microlenses on the sensor "miss" the light
sensitive area of the
pixel.  However,  because the microlenses increase the efficiency of
the light sensitive area by
a substantial margin, they are used.  However, at extreme incident
angles, the microlenses
don't work.  A Nikon 200mm f4 lens wide open produced a 45%
illumination loss at the
18mm (edge of the FF sensor) mark.  However, at f8, the loss was down
to 10%.
Obviously, with a sub-FF sensor (Nikon's own) the loss would be
miniscule.
The loss of the Nikon 200mm f4 was similar to a Zeiss 80mm at f2.8.
The three solutions to the problem are:
Reduce the aperture to around f8 especially with WA lenses that do not
support full field
illumination.
-Use larger format lenses. (This would seem pretty obvious, except the
speed loss is great)
-Use newly produced lenses designed to eliminate vignetting.
-They are made by a German company called;  Ingenieurburo Eckerl of
Hutthurm.
Loss at the edge with this company's lenses is about 2%.

http://www.ib-eckerl.de/lib/nav_lib.php?optik+0011