The Poisson cluster rainfall generation model has been underutilized in urban hydrological simulations due to challenges in reproducing fine-scale extreme precipitation. This study addresses this issue by comprehensively validating four variants of Poisson cluster rainfall models for three types of urban hydrological simulations: surface flooding, combined sewer overflow, and blue-green infrastructure performance. We generated 200 years of synthetic rainfall time series using the Base model (basic Poisson cluster model) and three variants: the Weight model (emphasizing variance and skewness in calibration), the Sine model (replacing rectangular rain cells with bell shapes), and the Weight + Sine model. This process was applied to rainfall data from 39 stations across Switzerland. Our key findings are: (1) While the Base model systematically underestimates extreme values, the Weight, Sine, and Weight + Sine models effectively address this bias. The Sine model directly enhances the intensity of fine-scale extreme rainfall, while the Weight model increases intensity at coarser durations; (2) Selecting the appropriate rainfall model depends on the timescales to which the urban systems are sensitive. For simulations of urban flooding, where fine-scale rainfall variability is crucial, models with bell-shaped rain cells outperform others. In contrast, systems like combined sewer overflows and blue-green infrastructure, which respond to coarser timescale variations, are better simulated by models using the Weight modification; (3) In all three simulation cases, the Base model failed to reproduce the quantile and interannual variability of hydrological metrics. This highlights the necessity of model modifications for reliable use in system design.