There is a slight difference (about 5 cm in 5 days) in filter and smoother predictions when using a ThrustHistoryFile in the filter and smoother prediction span. The filter and smoother predictions should be identical. They have much better consistency (although are not identical) when not using a ThrustHistoryFile.
The same difference appears in smoother ephems generated in a run that simultaneously generates a filter and smoother ephem, and one that only generates a smoother ephem.
See test case EKFS_SpacecraftNavigation_LEO_GpsPosVec_Fermi_FOGM_Cd_ADSF_WarmStart_PredDv.
Note to tester: See test case (DoNotDistribute) EKFS_SpacecraftNavigation_LEO_GpsPosVec_Fermi_FOGM_Cd_ADSF_WarmStart_PredDv.
Confirmed fixed in 2020_10_27 Fermi special build.