I thought I'd post the outcome of a little real-world field test we conducted. Mainly because the outcome was a little more interesting / unexpected than the test-bench / theoretical discussions up to that point suggested. Right up front if you're not into wildlife shooting at golden hour with long lenses, you can jog-on and correctly state that T stops don't matter and at best/worst cost you 1/3 stop.
What did we 'discover'? Well it looks like you can lose around 0.7 to maybe a 1 stop of light over the advertised F Stop on your lens when using a long zoom. In other metrics that would mean having to drop from 1/1000 sec to 1/500sec, or using a higher iso, or more interestingly for us walking away from golden hour maybe 20mins early (than using a prime at the same F stop and with the same camera).
Kit? We took a couple of cameras (to try different chips 1Dxiii and 5Diii) and four lenses, a newish Canon 600mm F4 is ii, a couple of Sigma 150-600mm f5.6-6.3 lenses (contemporary and sport), and an old Canon non IS 600mm f4. What did we do? Set everything to f6.3 and 1/1000 sec and watched what happened to iso on the camera displays over an hour or so as the sun set. The Canon's both performed "the same" as did the two Sigma's. The Canon's at f6.3 would chuck out an iso of 800 on the 1Dxiii and the Sigmas an iso of 1250 / 1600. From memory it was a similar differential on the 5Diii. We also played about with the DoF preview button just to make sure the Canon's weren't somehow sneaking in a bit more light with that f4 aperture.
If you arbitrarily set parameters after which you'd stop shooting, blurry images, too grainy etc. So for us we chose f6.3, 1/500sec, iso1600 (mainly because of the prevailing light) then in our part of the world you'd have walked away 20minutes earlier with the zooms vs the primes if everyone had an f6.3 aperture. Or put another way, if golden hour is an hour, you lose 1/3 of it.
Will my friend who started this ball rolling now go for the prime? Haha, actually no, she's going with the Sigma. Our little test allowed her to handle a range of different lenses all 'together' and explore those 'other factors'. T stop was more important than we were all expecting though.
For me I've probably come away liking my primes a little more, and probably (for my style of shooting) agreeing with DxO on the relative importance of light transmission as a factor for deciding between (long) lenses.
As a disclaimer I should say we don't plan on re-running this test, it's fairly time-consuming to organise 4 people with multiple lenses etc in the real world and we're not paid to do this. That said if we've obviously 'missed a bit' (that is likely to be material) please jump in. Certainly I think I learnt something new anyway, if we didn't jog the tripod when changing cameras, if a cloud didn't appear and lower ambient light levels as we changed camera, if..... All the best.