we are trying to compare something which reflects 18% of light to something emitting 82% of its maximum light.
Which to my mind makes the exercise entirely futile. You can make up any hex value you wish and then you can physically turn the monitor brightness (or backlight) up or down, as you like, so all in all trying to match an RGB hex value to 18% reflectivity seems like a nonsense to me.
What's the appeal with 18% reflectivity in any case? I thought modern digital cameras were calibrated to a reflectivity of 12.5% (some say 12%, some say 13%). But think about it this way....
In my experience, the headroom in a digital camera before clipping is 3 stops above a standard grey card exposure. i.e. meter off a grey card (or anything of even neutral tone, including white if you like) and then increase the exposure by 3 stops. You should just be nudging the clipping point, but not quite break through it. Now expose at +3.3 and you should see clipping appear within the central portion of the image. Light falloff at the edges will prevent the whole image clipping at once.
So, short story - you have 3 stops of head room above a centred meter reading. Thus, if you wanted to photograph something of 100% reflectance you would want to place that at +3 stops within your exposure. 50% reflectance would give you an exposure 1 stop below that. 25% reflectance would give you an exposure 2 stops below the clipping point and 12.5% reflectance would give you an exposure 3 stops below the clipping point. That is the aim of modern DSLR meter calibration, or so I believe. That is why the target reflectance of a grey card for digital is 12.5%, not 18%.
If you metered from an 18% grey card which is brighter than 12.5%, obviously, then the camera would "underexpose" your shot. Or, put another way, it would buy you an extra ~1/2 stop of highlight headroom. If you want 3.5 stops of highlight range then by all means meter from an 18% grey card, but I still don't see what that has to do with hex values for anything - or have I missed the point of this thread entirely?
Also, do we have to talk in hex? Is there something wrong with base 10, all of a sudden?
Incidentally, since gamma curves screw the tonal mapping of hex values, I wouldn't personally even attempt to guess what hex value equates to 18% (or 12.5%) grey, assuming that was even possible. On a linear tone "curve" you could assume that RGB = decimal 31,31,31 should be 3 stops below clipping, with 127,127,127 representing 1 stop below and 63,63,63 representing 2 stops below. None of that helps deal with turning the backlight brightness up or down though.
Well, that's what I think, anyway. I'll stick to metering reflected light or incident light. I can't see much use in caring two hoots about light emitted from some random monitor, unless you actually want to photograph that screen.
