The classic nonhomologous end-joining (c-NHEJ) pathway is largely responsible for repairing double-strand breaks (DSBs) in mammalian cells. XLF stimulates the XRCC4/DNA ligase IV complex by an unknown mechanism. XLF interacts with XRCC4 to form filaments of alternating XRCC4 and XLF dimers that bridge DNA ends in vitro, providing a mechanism by which XLF might stimulate ligation. Here, we characterize two XLF mutants that do not interact with XRCC4 and cannot form filaments or bridge DNA in vitro. One mutant is fully sufficient in stimulating ligation by XRCC4/Lig4 in vitro; the other is not. This separation-of-function mutant (which must function as an XLF homodimer) fully complements the c-NHEJ deficits of some XLF-deficient cell strains but not others, suggesting a variable requirement for XRCC4/XLF interaction in living cells. To determine whether the lack of XRCC4/XLF interaction (and potential bridging) can be compensated for by other factors, candidate repair factors were disrupted in XLF- or XRCC4-deficient cells. The loss of either ATM or the newly described XRCC4/XLF-like factor, PAXX, accentuates the requirement for XLF. However, in the case of ATM/XLF loss (but not PAXX/XLF loss), this reflects a greater requirement for XRCC4/XLF interaction.