I think I understand what you're saying, Omni . . . and if teams were better at making 2s I would agree with you completely.
(And, as I said, if a team made its 2s at greater than 49% proficiency -- or, if they have a particular play that they know will work most of the time -- then going for 2 should be the rule rather than the exception.)
But what your line of thinking just doesn't seem to sufficiently respect is this: every time the average team runs the average try for two, it lowers its expected total point production over the course of the game. When the teams each still have three or more possessions left in the game -- and the last possessions of the game are the most successful offensive ones at getting 6s and 3s, beyond any doubt -- then why do you want to lower your point production?
Especially when, in your example (as I understand it), it is not guaranteed and indeed maybe not even likely that the minority scenario (that is, you make the early 2) will reduce the number of scoring possessions that you need to win or will increase the number of scoring possessions that the other guy needs to win.
To me, if -- as is true most of the time -- there are enough possessions left in the game for the risk of having lower total point production to come back to haunt you (which would occur if you are going to end up ahead by only 2, 3, 6, or 7 or 8 instead of up by 3, 4, 7, 8, or 9, or you are going to end up behind by 3, 4, 7, 8, or 9 instead of by 2, 3, 6, 7, or 8: in other words, a LOT of scenarios) -- then the 2 seems like a bad risk to take.
The argument that an early missed 2 isn't so bad because at some point "you are going to need to make a 2-pt conversion to win" doesn't do it for me. Again, you are more likely to miss it than to make it. That's why (or should be why) coaches will not go for 2 to win the game in the last 20 seconds. They'd rather take the tie and put faith in the 50% coin flip and their D.
There's actually all kinds of debate about among math geeks (which I'm not at all, BTW), and no point of view seems conclusively proven. Your line of thinking has its supporters, and I was much too quick and simplistic yesterday in dismissing it.