To consolidate the safety assessment for liquid-metal fast breeder reactors (LMFBRs), hypothetical core disruptive accident (HCDA) sequences have been extensively studied over the past decades. Numerous analyses of the so-called initiating phase (or primary phase) of a HCDA have been made with the safety analysis system code SAS4A. The SAS4A accident analysis code requires that subassemblies or groups of subassemblies be represented together as independent channels. For simulating a severe accident sequence, a subassembly-to-channel assignment procedure has to be implemented to produce the consistent SAS4A input decks. Generally, one uses imposed criteria over relevant reactor parameters to determine the subassembly-to-channel arrangement. The multiple-assembly-per-channel approach introduces corewide coherency effects, which can affect the reactivity balance and therefore the overall accident development. In this paper, a subassembly-to-channel assignment procedure based on the subassembly power-to-flow ratio is presented and implemented to generate the SAS4A input decks over a range of parameter values. The corresponding SAS4A calculations have been performed on a large LMFBR. The purpose of the present series of calculations is to investigate the magnitude of errors encountered in the analysis of the initiating phase related to the subassembly-to-channel arrangement selection, by comparison with a one-subassembly-per-channel reference solution. It appears that a refinement in the channel arrangement substantially reduces corewide coherency effects. Analysis of the calculations also suggests that an accurate representation of the scenario requires the number of channels to be on approximately the same order of magnitude as the total number of subassemblies. Numerical results are examined to provide the reader with quantitative measurements of bias related to subassembly-to-channel arrangement.