Yes, you can quantify in any particular medium in which THC would be residual. For saliva, if I recall, our detection level was 15 ng/ml to simply report as a pos/neg result, but we could actually see it down 0.5 ng/ml on an optimal instrument, and report as a quantitated result per client request. GCMS limits could probably get even lower. Hell, I believe you can quantitate down 20 or so ng/ml on simple EIA instruments (the fancy version of the color changing immunoassay cups we've all used for jobs), which is fast as hell and requires zero sample prep, though it's been a while since I manned an Olympus.
As for intoxication determination, it shouldn't be too difficult to find subjects for the "smoke this cannabis and use this driving simulator" test. From there, they can determine at what level the average person is too intoxicated to drive safely.
The saliva test is taken as a cheek swab, so having a particularly dry mouth after burning some trees is what is helping give a more concentrated reading. Like I said, there is a very short detection window for saliva (and slightly longer in blood) for getting more immediate results without the holdover that would be present in urine.
I'm sure there are also screening confirmations that could be developed for any number of the other chemicals present in pot besides big, hairy assed 11-nor-9-carboxy-Δ9-THC that could be more beneficial in pin-pointing usage time.