In time-domain measurements of a Poisson two-level system, the observed transition rates are always smaller than those of the actual system, a general consequence of a finite measurement bandwidth in an experiment. This underestimation of the rates is significant even when the measurement and detection apparatus are 10 times faster than the process under study. We derive here a quantitative form for this correction, using a straightforward state transition model that includes the detection apparatus, and provide a method for determining a system's actual transition rates from bandwidth-limited measurements. We support our results with computer simulations and experimental data from time-domain measurements of quasiparticle tunneling in a single-Cooper-pair transistor.