As defined in this post, the monopolar EMG can reduce the common mode noise by taking the difference of the two inputs, i.e. EMG_mono = k * (V1 - V2). And, for bipolar mode,EMG_bi = k3 * [k1 * (V1 - Vref) - k2 * (V2 - Vref)].
To simplify the equations, we assume k1 = k2 = k3 = 1. Then monopolar is simplified as: EMG_mono = V1 - V2, and bipolar is simplified as EMG_bi = (V1 -Vref) - (V2 - Vref) = V1 -V2. That is to say the overall effect is the same for both monopolar and bipolar modes.
The question is: why is bipolar better than monopolar at reducing (common mode) noises?