One of the biggest challenges in optically stimulated luminescence (OSL) dating is to deal with sediments that are affected by post-depositional mixing. Equivalent dose (De) distributions of such samples are commonly analysed using the finite mixture model (FMM), which can statistically identify dose components of grains derived from sedimentary layers with different ages. The FMM has been increasingly applied in OSL dating of sediments from key geological and archaeological sites that are critical for understanding the evolution of Quaternary environment and human history. The burial age of the dated layer is commonly determined by dividing the De value of the target dose component estimated using the FMM by the total dose rate of the bulk sample. However, a systematic investigation of the reliability of this method for different post-depositional mixing situations has not been conducted. This study aims to investigate the reliability of the burial ages so obtained, using simulated mixed single-grain De distributions. We assess the performances of two different protocols—a user-specified or an automatically estimated between-grain dispersion value, σb. A number of factors influencing the reliability of the determined burial age are investigated, including the inherent dispersion in different dose populations, the proportion of grains associated with the target layer, the sample size used, the difference in dose rates and burial ages between the mixing components, and the time when mixing occurred. Published single-grain samples affected by post-depositional mixing are synthesized and their FMM results are assessed. Our results show that the reliability of FMM is controlled by the analytical and the intrinsic errors and depends critically on the proportion of grains for the target layer. The simulation and synthesized results caution against using an arbitrarily specified σb as input for the FMM. The intrinsic accuracy of the burial age is controlled by multiple factors for both synchronously and asynchronously mixed samples. The FMM ages derived from sediment samples affected by serious post-depositional mixing should be constrained by independent chronological information.