A foature common to predictor control shemes for plants with time delays is the use of a plant model. Any difference between plant and model is referred to as mismatch ; this introduces additional parameters which may be exploited to produce performance improvement. In previous contributions on this topic, the improvement reported has been by overestimation of the time delay. This paper contains several examples of the Smith scheme which are improved by underestimating the time delay. In each case, a useful approximation of the optimal model delay may be achieved by a technique introduced by Marshall and Salehi (1982). The potential performance improvement by mismatch depends on the plant delay. Consequently, the addition of the relevant time delay into the Smith sehome to increase artifically the plant delay may allow further performance improvement by mismatch. The required model delay is determined for two different locations of this additional time delay. The advantage of the less intuitive location is that the required model delay is always realizable. The performance of delay-free scheme is improved by the addition of a feedback loop containing two time delays.