You want the exposure mapping to affect the background, as it’s really part of the rendering when you’re using the maps.
http://www.chaosgroup.com/forum/phpBB2/viewtopic.php?t=16078&highlight=
The last post was a test I did using the sun/sky system. You really have to think of it in terms of real-life photography, and adjust the f-stop, shutter speed, and film ISO. For instance, on a sunny day most people would recommend 100 iso film… You’ll have to adjust all these settings as you move the sun around, a shot taken with a twilight sky will need a slower film speed and/or a slower shutter.
So, for instance, we’ll start out with the camera set at f = 8, shutter = 500, and film speed = 200.
Changing the shutter speed only x2 (from 500 to 1000) cuts the amount of light let in by 1/2 (making the image more dim).
Changing the f-stop only (from 8 to 11) does the exact same thing, reducing the amount of light by 1/2.
Changing the film speed only x.5(from 200 to 100) does the exact same thing, reducing the amount of light by 1/2.
In the opposite direction:
Changing the shutter speed only x.5 (from 500 to 250) increases the amount of light by 2.
Changing the f-stop only (from 8 to 5.6) does the exact same thing, increasing the light by a factor of 2.
Changing the film speed only x2 (from 200 to 400) does the exact same thing, increasing the light by a factor of 2.
You could probably walk out on a clear, sunny day… put your camera in fully auto mode, record what it recommends for the exposure, and plug those into the camera and they’d be pretty close.
Also make sure you have a ground plane, I think that helps with any stray GI from the sky map. Incidentally, if you’re using linear workflow, 128-128-128 greyscale is pretty light and gets blown out on tests I do, even when the lighting looks correct with actual materials applied… you might want to make a bitmap that’s 128-128-128 (at gamma 2.2) and use that on the material as a test material.
On second thought (after writing this whole thing) looking through your post again it’s probably is a problem with using linear color mapping… my test on the other page was done using linear workflow, with Gamma Correction color mapping and inverse gamma set to .454