These results are what one expects from a cubic crystal and a twobody
potential. It is not impossible the ReaxFF potential for Au behaves
*locally* as a pair-wise potential, but it seems unlikely. Have you
checked to see if you can reproduce properties from the paper?
ReaxFF potentials have rough potential energy surfaces. Your results may depend on the strain rate used to compute the elastic constants. You may want to try the calculations varying your strain rate from say 0.0001 to 0.002 in 5 equal divisions.
You can then plot each C_ij elastic constants vs strain rate to see if you get something close to the published data for any one of the strain rates. For metals generally, a strain rate above 0.002 start to get beyond the elastic limit.
Suleiman.