Turbidity currents transport fine sediment through reservoirs, with their propagation influenced by changing inflow conditions and sediment exchange with the bed. This study investigated the first turbidity-current event during the 2004 Xiaolangdi water–sediment regulation operation using a two-dimensional layer-averaged model calibrated against field observations. Calibration against observed front positions yielded a root-mean-square error of approximately $1.8\mathrm{km}$ and a Nash–Sutcliffe efficiency of $0.90$. The simulated current evolved from an initial erosion-dominated stage to downstream deceleration, deposition, and attenuation. Stronger hydraulic forcing, higher sediment concentration, and longer high-intensity inflow generally promoted downstream advancement. Varying the distribution of inflow altered the propagation process and endpoint advancement under equal cumulative water volume and sediment mass, identifying input timing as a control alongside total supply. The two stratified beds comprised equal-thickness layers in opposite vertical orders and shared the same combined grain-size composition as the uniform reference. These configurations changed $24\mathrm{h}$ front advancement by less than $2\%$ relative to the reference and modified local velocity and response timing. Simulated grain-size distributions revealed spatial variations in sediment sorting. Together, these findings support joint evaluation of inflow forcing, bed-sediment availability, and sediment-exchange parameters in reservoir-scale modelling and process-based assessments of sediment routing.