In my opinion, if pairing are random, results should be influenced by that random factor.
eg. high skill player vs a random noob => the high skill player should win, because he have more skill and was paired with a random lesser skill player => hence that result of that 1st round is more due to a random factor than an actual well-made, acurate and fair pairing. And those happens in last rounds.
Well, assuming the national champion will go high ranking in that tournament, someone who get to play him in round 2 allready won his round 1 ? So that round 2 opponent have more legitimity than the round 1 lesser skill opponent the nat champ had ?
In round 3, players get to play better level opponents, or worse. That last round you made vs players of your skill is more important, to me, than the game you made 2 rounds ago vs players nobody could say they were of your skill or not.
And the more you get rounds, the more precise everybody know if you were doing well in a tournament or not, the less random pairing are and the more fair they get.
So the more important last round should also weight in the SoS than the previous, to me.
Because pairing are less and less random the more you get rounds.
Hence that progression of weights.
Granted, you lost vs the local netrunner champ in low rounds. But what is important to show your real strength to break a tie with players of your skill is how you played vs players known of your skill, not of those of skill too far from you that everybody knows you must win or loose.
It's not focused on smoothing, because to me smoothing does not help to break ties.
It's "legitimazing" the SoS.
For a 4 round tournament, if I choosed (0,1,2,3), counting 6 scores, at the end I have more dispersion in players' SoS than a regular (1,1,1,1) counting 4 scores.
If I choosed (0,1,1,1), counting 3 scores, then I have less dispersion than the actual 4 scores SoS (and so, more smoothing).
If really no-round-1 is a problem (it is not, to me, because round 1 is throw-a-dice-get-your-points, but I understand...), we could talk only about (0.5, 1, 1.5, 2), (1,2,3,4) or a fibonacci suite (1, 1, 2, 3, 5, 8, 13, etc).