There are a number of considerations in the measurement of light.
Light is a form of radiant energy. In the case of photographic light meters, the spectrum of sensitivity of the meter is important. Ideally, it would correspond with the Spectral Sensitivity Curves of *ALL* film. The response curves would necessarily have to be limited to spsecific spectral areas, and would match the sensitivity levels at those areas.
When we speak of a film - or meter - "balanced" for 5500K, we are describing a spectral curve that approximates (best effort basis) the spectrum and light levels of "Noonday Sunlight at 40 degrees North Latitude - etc.", not an extremely narrow band at only one single frequency - laser-like.
There are lots of compromises: It is extremely difficult to get the meter to have the same sensitivity levels as sunlight in the first place. Films differ widely in their sensitivity curves (example: look at the spectral sensitivity of Infra-red film), and in particular, the light measured, in itself, is more than likely NOT have the same spectral composition as Daylight at 40 degrees.. etc.
Then - added to all the "compromises" we must include - metering technique - Incident or reflected? When we use "incident" to we deliberately point the sphere at the source of light, or the camera itself? How does one "point" a sphere anyway?
Fred Picker of Zone IV (did I get that right?) tried to eliminate some of the compromises by adding filters to Honeywell 1 Degree Spotmeters to respond more closely to the spectral sensitivty of average black and white films. Great, but that probably diminished their accuracy with color films.
In the last analysis, we probably do not deserve to get exposure as close as we do - even with the fancy-shmancy 10,320 point matrix blue-pont sorcerer's dust compensated systems..
I suppose a truly accurate meter could be built. Sensors would analyze the light intensity and specrtrum of average and specific points on the subject: that information would feed and control various (and many) filters and match that information to programmed Film Spectral Response curves which would be further processed to respond to any correction filters on the camera - and any biases caused by lens coating.... and, whoops, nearly forgot... some super-intelligent future Processing Prediction software.
There! Sounds like everrything. Now to make the thing the same size as the average meter and keep the price under a few hundred dollars. And then, SELL it to a bunch of "Digidumbos" whose greatest acheivement so far is figuring out which end of the camera to point at the subject....