With this in mind, I would trust information from this sensor - MUCH MORE than I would put faith in my own perception - remember, IR can NOT be seen.
It's not my perception. I have cameras with TTL meters that I have compared it with. Under multiple circumstances I have verified that the silicon sensor is just wrong, because of how it departs radically from what other exposure meters (which have proved to be working and accurate) indicate, for certain targets which are known to emit (reflect, whatever) IR well.

Like I said in my original post, by reading the TLR2030 datasheet, I didn't think spectral response would end up being a problem, but it clearly is. I will consider the problem solved when the TLR2030's output tracks reasonably well with other exposure meters that are proven accurate, instead of departing wildly for certain targets.