The DIRAC calculations imply the local Density Functional in its simple X-alpha approximation without any gradient corrections. Not the scalar relativistic but the fully relativistic Hamiltonian is used, including spin-orbit coupling. In ADF you may use the scalar relativistic Hamiltonian and most users will employ a more sophisticated lda than X-alpha, such as the default vwn (Vosko, Wilk, Nusair) formulas, and may in addition routinely apply gradient corrections. The core potential may not exactly match the Fock operator applied in the molecular calculation. The effect is very small and one can neglect the discrepancy.
