Higher k_eff at 0 depletion step


I’m observing some strange behavior where the 0 step depletion k in OpenMC is higher than the eigenvalue simulation:
1.03869058e+00 +/- 3.76492873e-04 (depletion) vs. 1.03693 +/- 0.00021 (eigenvalue)
For comparison, SCALE is giving:
1.03683 +/- 0.00023 vs. 1.03701 +/- 0.00019
I have checked the material.xml file atom densities at step 0 with the SCALE output values and they appear to be the same.
Graphite moderated Molten Salt (FLiBe w/~20% enriched U) reactor core. Using CECMIntegrator.
Any ideas?