The Test
I took two sets of three images of a white target.
The two sets were taken with

1. Normal light
2. Polarized light

1. Without polarizer
2. With polarizer set to minimum effect
3. With polarizer set to maximum effect

The Results

	Min		Max
Normal	-1 2/3		-1 2/3
Polarized	-2/3		-10 5/6

Analysis
As expected, in Normal light; the polarizer was a consistent 1 2/3 stop penalty.
As expected, in Polarized light; over 9 stops more light is blocked when the polarizer is adjusted to 'maximum effect'.

The suprise is that at 'minimum effect' the polarizer let 1 stop more polarized light through than normal light.
On reflection (no pun intended) the explanation seems "obvious".
Normal light is comprised of light at all polarization orientations; so even at 'minimum effect' the polarizing filter blocks more normal light than polarized light that is in "perfect" phase with the filter.

It's interesting to note that the 1 stop difference between Normal and Polarized at 'minimum effect' is 50%.
So it is as if the transmittance penalty is 2/3 stop for the filter and 1 stop because of the random polarization of normal light.
(For a total of 1 2/3 stops for this filter)

Bill
(edited to fix math error)

Visit me, info and galleries at: Nikon Photos and Information

Yup. That's what I would expect.

Set at the right angle, the polarizer itself matched that of your polarized light source, thus passing as much light as possible (limited only by the filters actual density). By contrast, randomly polarized (ie: unpolarized) light would always be at least in part blocked by the angle of polarization from the filter.

Bob Johnson - Earthbound Light - Nikonians Gallery
Nature Photography from the Pacific Northwest and beyond
Weekly Phototips and Articles - Lots of images

The 'actual filter densities' (as Bob puts it) were 1/2 stop and 2/3 stop respectively; similar to the 2/3 stop result in the inital post.
(The linear doesn't have the quarter wave retarder so maybe that's why it does slightly better.)

Also, the additional transmittance penalty remained at 1 stop.
(I wasn't sure the linear would behave the same way as the circular, but it did.)

Finally, these two filters were even better at blocking polarized light with results of over 11 stops as opposed to 9 in the original post.
(Not that it matters, 9 stops is plenty of attenuation!)

Bill

Visit me, info and galleries at: Nikon Photos and Information