Hello Yuan,
sorry for the delay, I thought there was something wrong with the way you write the file (no need to open units, and also there was a weird closing of the output unit in a separate entry of MGDRAW). Regardless, this is not the issue.
The main issue is that ENDRAW scores energy losses to a material, but neutrons do not interact electromagnetically, and the interactions they undergo result in nuclear recoils or other secondaries. These particles are then transported, and thus are not counted as losses, just energy transfer. A user had a similar question last year in this post. The ENDRAW is then invoked as charged secondaries slow down in the material, depositing energy in a form that is not transported (below delta-ray production, e.g.)
I hope this clarifies the issue,
Stefano