Large discrepancy of neutron current between Geant4 and MCNP in low-energy range (<=0.19MeV)

Hello everyone,
I use the Geant4_10.6 and choose the Physical List “Shielding.hh”( “G4NEUTRONHP_DO_NOT_ADJUST_FINAL_STATE”). I use the neutron data libraries, ENDF-VII.1, In order to be consistent with MCNP6.
The model is simple. A 14MeV monoenergetic neutron source and a sphere sensitive detector encloses the neutron source. All of these in a tub which is full of a kind of element, like Si. And I get the information of Neutron Current spectrum from the detector.

The issue is that Geant4 is giving me the following spectrum,

The relative difference is defined, R=|C_G-C_M |/C_M
In the neutron current spectrum 93% of the channels have relative differences of less than 10%. The greatest differences are mainly concentrated in the first four channels: 0-0.19meV.

But the official comparison documents I have seen show that the neutron spectrum varies significantly only in the <1eV range.

image

So my question is why low energy neutrons in Geant4 cannot match MCNP6. Is there some explanation? Are there ways to improve their matching?

Thanks in advance for your help!
Best regards,
Yi Ge
(I am new to the forum so I hope everything is all right with this post.)