Molecular dating often struggles to align with fossil records when divergence times are estimated in deep evolutionary history, particularly during the Palaeozoic era. This discrepancy may arise because of the assumptions embedded in constant birth-death rate priors, which oversimplify complex evolutionary dynamics and remain the default setting in most analyses. In the present study, we tested the impact of different models and priors on phylogenetic inference on a dataset of Rhynchonellata (Brachiopoda), and explored whether the divergence times inferred from molecular clocks and fossils can be reconciled. In particular, we addressed rate heterogeneity, that is, variation in speciation (lambda) and extinction (mu) rates over time, in divergence time estimation by incorporating the birth-death rates inferred independently from fossil occurrence data as rate priors and calibration points. These were reviewed using Marshall's temporal calibration method and ancestral state reconstruction. We showed that using the birth-death skyline model as a tree prior significantly reduces the gap between age estimates. Furthermore, to assess the generality of these patterns beyond a deeply diverging clade, we conducted a robustness check using Lingulidae, a comparatively young, re-diversified lineage, and determined that skyline models also produced divergence-time estimates more consistent with those inferred from the fossil record in this dataset. Overall, we showed that using a constant birth-death rate prior, which ignores rate heterogeneity, can affect the accuracy of the resulting phylogenetic trees. Our results revealed that when the lineages used for molecular dating persisted through the Cambrian explosion, Ordovician extinction, and Permian-Triassic (P-T) boundary, caution should be implemented in the interpretation of phylogenetic inferences as the data from molecular alignments alone and standard model assumptions may not reflect the significant shifts in diversification dynamics that occurred during these time periods. These findings highlight the inadequacy of constant birth-death rate priors in the presence of rate heterogeneity across deep time, thus limiting their reliability for studying lineages with complex evolutionary histories. Our analyses showed that informative priors from independent analyses of the fossil occurrences can improve the quality of molecular clock analyses, indicating the requirement of more flexible models to improve the precision of molecular dating in deep-time research.