Yes, 2…b5 is an unlikely move. But how does the gameplay compare between 2…b5 and 2…b6 – let’s analyze the score vs move with Stockfish 5:
First the actual game score (well, the actual moves published widely on the net), with 2…b5:
(Click images to enlarge)
Then the FSR suggestion with 2…b6:
There is little question that 2…b6 makes the game more consistent. On the other hand, the actual evaluation tracks between the versions, never really straying more than 0.5 pawn units. This suggests that the 2…b5 version isn’t overwhelmingly compelling.
To change a scoresheet, to me requires a high little of proof. I would say it’s somewhat like the different burdens of proof used in a court of law. Yes, the preponderance of the evidence suggests 2…b6 is correct (really, who could argue with that?). But to actually change the scoresheet, requires evidence beyond a reasonable doubt.
For whatever reasons, the actual gameplay does not meet this stricter burden of proof. And so I would judge the scoresheet not guilty, but allow an appeal to a higher court (i.e. a majority consensus of biographers).