I would suggest there is no correlation between visible light and IR in the atmosphere.
I don't think that's right. After all, the sun's output is more or less consistent at some relation of visible to IR---the difference in that proportion between different times of day is just down to differences in scattering and absorption. As we all know from absorbing AA's _The Negative_ with our mothers' milk, longer wavelengths scatter less, which is why (1) red filters cut haze and (2) sunsets look reddish.
So at a clear, cloudless midday, the amount of IR is low relative to the visible, because not so much visible light is getting scattered; in hazy golden-hour light, the proportion is high. At a given location, I think it's possible to work out metering and compensation practices for these cases and interpolate between them.
By the way, I've also had some luck metering through an 89B (mainly in sunny conditions and temperate latitudes). The meters in question may just be reading off the little bit of far-visible-red that gets through the filter, or they may be sensitive a bit beyond the visible spectrum---I don't know why it's worked but it has.
-NT