It is clearly shown that due to the asymmetry in the exact distribution, alphas
calculated with control limits using the normal approximation, suffer a significant
accuracy with regard the alpha value preset when the true proportion of
nonconforming is small or the sample size is not big enough.