Or would it be a 25% error, i think it would make most sense to calculate the error-difference in relation to the actual value instead of in relation to the erroneous value.