This paper provides a generalization of certain classical Fourier convergence and asymptotic Toeplitz matrix properties to the case where the underlying orthonormal basis is not the conventional trigonometric one but rather a rational generalization which encompasses the trigonometric one as a special case. These generalized Fourier and Toeplitz results have particular application in dynamic system estimation theory. Specifically, the results allow a unified treatment of the accuracy of least-squares system estimation using a range of model structures, including those that allow the inclusion of prior knowledge of system dynamics via the specification of fixed pole or zero locations.