Cool project, for fun at least!
Fun??!! I've resoldered and traced point-to-point 22ga solid core wire hundreds of times and written 700 lines of C. I'm doing this because I need a spot meter!

In the meantime, here's another idea. If you have an IR-pass filter (visually opaque), you could take two readings, with and without filter. The difference between "everything" and "IR only" = visible light.
That would work and be a good diagnostic, but it wouldn't be acceptable for the meter to actually work that way and have to take 2 readings or use 2 chips. I'm aiming to have a usable meter here; I already have EV, aperture priority, shutter priority, and film speed adjustments coded and printing on a 2x16 character LCD panel. I don't have any fancy averaging routines, but the meter lets you shoot the target and cycle through applicable camera settings

"Both LOOK ... ". Not relevant.
Of course it's relevant. If two things appear to be the same tone but one of them measures several stops brighter then you know something is up. Especially when it's something known to emit lots of IR (grass).

Do you have information about the sensitivity (@ given wavelength/s) of your IR cell?
Yes, and using a complementary colored filter to "even out" the red-leaning response would be a good idea, but I don't know where to find color gels or information on their transmission spectra.

In the meantime I dissected my wife's old cell phone and pry'd the IR filter off the CCD. Unfortunately I broke it getting it out, but the biggest piece barely covers the sensitive area on my light sensor. I'll have to wait till tomorrow to recalibrate and see if the IR anomolies are still present. Surprisingly the IR filter sucked down almost a stop under mixed CFL/incandescent interior lighting.