I purchased a red 25 filter for outdoor black and white pictures hoping that
it would yield better results than just using the red channel of a full
colour capture as it would probably get a fuller exposure and therefore less
noise and grain.  However what I find is that it is very easy to
oversaturate the red channel of the resultant image.  Is there a significant
difference in colour sensitivity between the image sensor and the camera's
light meter?  Is it a problem in de-bayerizing?  The green and blue channels
still seem to get a fair bit of data and the histogram shows a large
underexposure (about 4 stops) even though the red channel is saturated by
about one stop too much.  The camera is a Canon Rebel XT (350D) and of
course I am capturing RAW.

I also purchased a non-coloured neutral density gradient screw-in filter.  I
am not really pleased with the results yet.  It is much more effective at
wider angles for obvious reasons, but since all of the effective lens area
is used for forming the image from rays striking it at acceptable angles,
then it would seem that the filter should need to be spaced off the lens
significantly to be effective.  Stacking a few UV filters between the lens
and the ND grad filter seemed to make it more effective at longer focal
lengths, but the stack of UV filters leads to noticeable flare and of course
then there is vignetting at wide angles.  I know that the purpose of the
filter is to reduce the dynamic range of the shot where you can divide the
brighter and darker portions of the picture can be roughly divided in the
middle.  Just so far the images are not spectacular and not really better
than without the filter.  It appears that this is not a simple filter to use
but I am willing to make an effort.

The question is how do you use a ND grad filter effectively for good results
and where to meter?  Do you try to meter the middle area or upper or lower
portions?  Matrix metering seems to result in over-exposure which makes
sense, center weighted averaging might make sense and partial metering
probably would work but still it seems to be necessary to bracket the
exposure.  Would it make sense to meter off of an 18% grey card held
opposite the light source?

Thanks for anyone's insights or experiences with using these filters with
digital capture.