I have to say that this possible effect has caused me some confusion regarding my understanding of throw and how to predict it, so I did some measurements to see what is actually happening.
My test light was a UF-1504, with 62mm diameter lens and XPL HI V2 1A. I put in a 8x7135 driver to keep the output more constant with time. I measured the lux at three different distances measured from the lens.
dlux__cd
6.05m7230_264.6Kcd
8.34m3840_267.1Kcd
10.29m2530__267.9Kcd
These measurements are consistent with measuring the distance right from the lens and not at some point behind the lens.
Plugging the 10.29m measurement (the beam size at this distance was 36cm) into the calculator you linked (after converting units), results in:
divergence distance behind aperture: 2.14m
candlepower: 390.9Kcd
The calculator results are not consistent with my measurements and how they vary with distance. There are certainly some things in your explanation of the effect that make sense, but actual measurements tell a different story.
A different effect could have caused your conflicting results when you measured the lux at 1m and 2m. At small distances like 1m, the lens might not be filled with the image of the LED, from the point of view of the lux meter. Moving farther back to where the lens is filled would then result in a larger throw number.