This paper shows that a result recently obtained for arbitrarily improving the zero-state transient performance of a continuous-time model reference adaptive control scheme, does not have a counterpart in discrete-time. Using standard l(1) optimal control theory, it is shown that unlike the continuous-time case, it is not possible to modify the control law to obtain arbitrarily small l(infinity) gain for the closed-loop function from the parameter mismatch to the error. Nevertheless, the control law may be modified to minimize this gain, perhaps resulting in improved error transients.