It's not terrible, but the simulation is correct, it's your equations that are incorrect. A simple way to look at it is that the 1.25V REF is, in effect, your common reference for the differential amp connection, which gives it a gain of +1 to the output. This offsets the output by 1.25V. Thus the output voltage is simply the differential signal gain of the two input signals plus the offset, or (V1-V2) * (R7/R3) + 1.25v = 3.67V, close to the 3.66V you simulated.