Well, yes.
That should get you 90% of the way there.
But you will always need to define what the 'weather' is like... which is what mie/ray/extinction_factor/clouds are essentially doing.
I'm happy to generate TOD curves that are specific for people. I'm currently using 3DS Max IES sky for the same date/locations as Racer and then measuring values from it every 1hr and then every 5 mins at dusk/dawn for added quality.
IES is used in the industry for architectural analysis of buildings and light impacts, so it's good quality data as a base to work off
But you also need to do two sets, one for a clear day and then one for cloudy (where diffuse is reduced and ambient is boosted)
Again, ideally, these would be generated as we go from clouds 0 > clouds 1 we would see the sun become occluded (lower it's intensity in the live envmap and thus lower the intensity of the diffuse value), and ambient to diffuse value ratio would increase (clouds become more illuminated than natural sky background colour in some cases, less blue tint in envmap etc)
I think currently Ruud is using a mipped envmap to determine scene intensity for the auto exposure, soooo, in theory, if we also sample the envmap mip when it's at say 16x16, we should get a good idea of the intensity point from the sun (take brightest value from the mip), and then take the average value of the envmap top half (sky only), then try work out the amount to subtract from that which is due to the sun influence (sun intensity in rgb divided by pixels in top half of envmap?)
Easy peasy... in theory... it's getting the maths right... but Ruud is good at maths
Dave