The problem of improving the stability characteristics of discrete-time SISO reduced order adaptive controllers is addressed in this note. A method is proposed to adjust the observer poles of an indirect adaptive scheme in order to ensure stability of the closed-loop system in spite of possible unstructured uncertainties (e.g., process-model order mismatch). Assuming that the identified model transfer function module converges within a previously defined uncertainty range, the frequency response-like compensation method is shown to lead to a stable design, provided that the reduced order estimated model is stable and stably invertible and the system is open-loop stable.