Conductor fill is calculated based upon the cross sectional _area_ that the conductors take up in the conduit. However once you have a very few number of conductors nearly maxing out a conduit, you need to consider the linear _diameter_ of the conductors.
Rather than dealing with the proper amount of open space needed for conductors to freely move, consider the area of conductors that just barely fit in the conduit. I'll also ignore 'trade sizes'; the hypothetical 1" conduit that I am describing has a real internal diameter of 1"
Start with a conduit that is 1 inch in diameter. This has a cross section of 1000000 circular mils. A _single_ conductor that just _fills_ this conduit would also have 1 inch diameter, and thus a cross section of 1000000 circular mils.
Now look at two conductors in this same conduit. These conductors would have to have a diameter of 0.5", because two side by side would exactly fill 1". The combined cross section of these two conductors is 500000 circular mils. Thus for _two_ conductors, just barely packing a conduit full, the maximum cross sectional area is only 50%
Now look at three conductors in the same conduit. These conductors would have a diameter of 0.464", and three laying in a 'triangle' would just exactly fill the 1" conduit. The combined cross section of these _three_ conductors is 646171 circular mils. So with three conductors and maximum packing, you get to 64.6% fill.
The ratio of 50 to 64.6 is equal to the ratio 31 to 40, give or take a bit of rounding. Unfortunately this doesn't quite work for a single conductor, where keeping the same ratio would permit 62% fill, but code only permits 53%.
-Jon