We investigate the difference between analytic predictions, numerical simulations, and experiments measuring the transmission of energy through subwavelength, periodically arranged holes in a metal film. At normal incidence, theory predicts a sharp transmission minimum when the wavelength is equal to the periodicity, and sharp transmission maxima at one or more nearby wavelengths. In experiments, the sharpest maximum from the theory is not observed, while the others appear less sharp. In numerical simulations using commercial electromagnetic field solvers, we find that the sharpest maximum appears and approaches our predictions as the computational resources are increased. To determine possible origins of the destruction of the sharp maximum, we incorporate additional features in our model. Incorporating imperfect conductivity and imperfect periodicity in our model leaves the sharp maximum intact. Imperfect collimation, on the other hand, incorporated into the model causes the destruction of the sharp maximum as happens in experiments. We provide analytic support through an asymptotic calculation for both the existence of the sharp maximum and the destructive impact of imperfect collimation.