Possible luminance meters I've found so far

The issues

Sensitivity: According to Mark Fairchild's tech reports (scroll down), the luminance of white on a CRT he tested was 56 cd/m^2 (but he mentioned to me that this was lower than its maximum possible output), and on LCDs, he measured 153, 162, and 188 cd/m^2. I want to characterize the output curves for the individual primaries, and the hardest one will be blue, since its less luminous than red and green. The middle row of RGB to XYZ conversion matrices usually puts blue's contribution to luminance at about 10%, leading to luminances of only 5.6 cd/m^2 for CRTs and about 16 cd/m^2 for LCDs.

If I (somewhat arbitrarily) decide that I need to resolve 300 luminance levels between the color being off (input level = 0) and the color being all the way on (input level = 255) that means I need resolution of about 0.019 cd/m^2 for CRTs and 0.053 cd/m^2 for LCDs.

Illuminance vs Luminance: A sensitive illuminance meter could suffice for this project. On LCDs, I wouldn't want the patch on the screen to subtend more than about 1/10 or 1/8 of a steradian, because then the view-dependent effects of the display could start kicking in. Combined with the fact that I want to see changes in low levels of blue, this puts my sensitivity requirement at about 0.0053 lux or 0.0005 footcandles (for a 1/10 steradian patch)! However, if I'm using an illuminance meter, I could put a diffusing surface on directly top of the screen, which would average out all the view-dependent effects, and allow me larger patches. But I'm not certain if averaging out view dependencies is a valid thing to do ...

Price: Under $500 would be best, but I'm curious what's possible for under $700 or so.

CRTs: CRTs have very non-constant luminance: as the electron beam sweeps by, the spot being refreshed is very bright, and then the phosphors die off pretty quickly. This page says it can get as high as 10,000 cm/m^2, and describes the risk of the device being inaccurate because of clipping at high values, as well as the need for long time-averaging.

Renting: Renting an expensive device is a possible alternative to buying a cheap one. For example, I found a company that can rent Tektronix J17 photometer (discontinued in the mid 1990s) with a luminance or colorimetry head for $344 a month. Renting a Minolta (see below) would be nice, but I don't know where.

Who else cares: A community of people that deals with these measurement issues is perceptual psychologists working on the human visual system. Using CRTs for producing test stimuli is cheap but the device has to be calibrated. Many of the companies listed above are mentioned on this page compiling one person's query and responses to a question about available devices. Also, David Brainard's Psychophysics Toolbox website has a similar listing of companies and distributors, but I don't think all of these are useful for pychologists.