Constant residuals commonly represent unresolved components of reservoir water balances, although their combined effects may vary with storage. This study investigated whether beginning-of-year storage explains annual water-balance residuals at Seymareh Dam, Iran, using inflow, outflow, and storage observations for 2011–2023. A linear storage-dependent residual was fitted using ordinary least squares and incorporated into the annual continuity equation. Parameter uncertainty and equifinality were examined through 10,000 Monte Carlo simulations within the Generalized Likelihood Uncertainty Estimation (GLUE) framework. Initial storage explained 32.3% of residual variability (R² = 0.323, p = 0.043). Compared with the no-loss baseline, the storage-dependent model improved in-sample Nash–Sutcliffe efficiency (0.5973 versus 0.3832) and reduced root mean square error (177.10 versus 219.17 MCM), although mean absolute error increased slightly. Using an NSE threshold of 0.50, 772 parameter sets (7.72%) were retained as behavioral. The resulting parameter-induced 95% prediction uncertainty envelope enclosed 9 of 13 observed annual storage changes (69.2%). Behavioral slope and intercept parameters were inversely correlated (r = −0.858), indicating equifinality. Leave-one-year-out refitting preserved the positive slope but did not establish independent predictive validity. The formulation offers a parsimonious diagnostic tool for long-term reservoir water-balance assessment, without resolving seasonal dynamics or individual physical loss processes.