Quantifying root water uptake (RWU) using hydrogen and oxygen stable isotopes and Bayesian mixing models is widely applied in ecohydrology. But its reliability remains unclear due to multiple uncertainties and the underdetermined nature of the problem. Here we systematically evaluate error propagation in MixSIAR-based RWU estimation using synthetic datasets with known RWU profiles. Through 2916 controlled simulations, we assess the contributions of eight factors, including observed soil and plant water isotopic compositions, RWU patterns, and model configuration (source grouping, prior information, and error structure) to the variance in accuracy and uncertainty of estimated root wate uptake proportions using MixSIAR. Results show that prior information and source grouping dominate RWU inference, explaining over 40% of the total variance. By contrast, when evaluated using posterior-predicted plant water isotopic compositions, their contribution falls below 10%, while soil and plant isotopic compositions and error structure together explain over 40%. This shift results in two things: (1) it masks the influence of prior information and source grouping for inferring RWU proportions using MixSIAR, and (2) it reveals a critical limitation in the approach: accurate isotope fits of observed plant water do not guarantee correct RWU inference, since multiple RWU solutions can produce identical isotopic compositions. These findings demonstrate that prior information and source grouping should be explicitly specified and carefully justified in future applications of MixSIAR in studies of quantifying RWU using Bayesian mixing models. Also, the multi-criteria evaluation frameworks beyond posterior predicted plant water isotopic composition are needed to assess the performance of MixSIAR in isotope-based RWU estimation.