Hi Siddharth,
Etiquette requires that we post discussions pertaining to problems with LAMMPS on the general mailing list. So my initial response was sent to you alone in error.
Meanwhile, if you are computing heat of mixing of binary/ternary alloys as a function of composition and getting wide variations from the same input file, then two things might account for this problem
(1) Your alloys have not reached configurational equilibrium at each composition.
(2) You are not taking enough configurational averages, assuming problem number (1) is not the issue.
For any thermodynamic property of a binary/tenary alloy at finite temperature, the alloy must reach configurational equilibrium before you can sampled for the desired property.
Suleiman.