Have you tried to to the same for your "Gain2" port? A simple check yields: >> 20.02 - 0.1/(0.5*0.01) - 0.01^2/(0.5*0.01) >> ans = -4.2674e-16. Perhaps this is what throws you out of tolerance.
Another thing I can recommend is trying with a different circuit. Nonetheless, I do agree with you that the tolerances ought to be looser. You may suggest it to Guy and Seth: http://blogs.mathworks.com/seth/2015/08/07/modeling-and-simulation-challenge-in-cody/

Go to Matlab and write:
>> format long;
>> 60/(2*pi)
ans =
9.549296585513721
That is the conversion constant of rpm to rad/sec. Using |format long| gives you the answer within Matlab's working precision.