We investigate the influence of mask duty cycle (DC)—the proportion of active ("on") pixels in binary-coded masks—on reconstruction quality in Snapshot Compressive Imaging (SCI) systems through comprehensive numerical simulations. Masks with DCs ranging from 5% to 80% are evaluated across two widely used SCI modalities: CASSI (Coded Aperture Snapshot Spectral Imaging) and CACTI (Coded Aperture Compressive Temporal Imaging). To bridge simulation and practice, we propose a realistic forward model for SCI that accounts for non-ideal system conditions, including image degradation, mask blur, sensor noise, and spectral discretization. Contrary to the common assumption that a 50% DC yields optimal performance, our results reveal that sparser masks (with 10–30% DCs) consistently improve reconstruction performance, particularly under high compression and high noise conditions. We further show that this range represents the practical optimum: higher DCs introduce degraded temporal sharpness or spectral fidelity, while lower DCs lead to reduced spatial resolution due to insufficient captured information. These findings offer practical guidelines for optimizing mask design in real-world SCI deployments.