In general, this is what is done with CPU heatsink testing when the results need to be compared despite varying ambient temps. It's the "rise above ambient temp" that is most often used for comparisons, which forms the basis of "C temp rise per watt" ratios. What Larry has done here is just an application of the same concept to case cooling.
You ask if we are really, really sure
-- maybe not that certain, we're always open to persuasion to valid counterpoint data or arguments. This is mostly science, after all.
Alright. I can't provide data points (might check during the coming week if I get bored), so let us keep this theoretical:
A simple technical argument
If the outside temperature was lower, it is quite likely that those fans would simply spin slower. But this would not actually lower temperatures (yet noise).
Especially in the case of the dual GPU configuration, this would be a significant effect. The same argument also holds for the CPU fan if it is temperature controlled.
Now, if the temperature of GPU&PSU are similar, the rest of the system will have different temperatures, but they will most certainly not be linear related to the outside temperatures.
A rather pedantic, simple theoretical argument
The heat sources are not equivalent, because at different temperatures we will have different amounts of losses. This is probably a negligible effect.
A slightly complicated physical argument
Presume the fans are spinning with the same velocity and the heat soruces are equivalent. The case under load shall be in thermal equilibrium, which means Heat_out = Heat_generated.
In a simplified version of reality, Heat_out consists of warmer air exhausted through the back and cooler air taken in through the front.
The argument is now, that the heat "lost" through cooler air taken in is only part of the total heat loss. Thus a change in this part of the system will not change the system as a whole in the same way.
A slightly complicated mathematical argument
If we restrict ourselves to the inside of the computer case, changing the outside temperature will not change the boundary conditions of the heat equation(s) in a linear way, because there is a more or less localized intake/exhaust point. Even if we presumed the simplified heat equation for a homogenous, isotropic medium, the resulting solution will not be simply scaled because the boundary conditions changed inhomogenously.