Hi yunzhen,
The reason is that such "fake" carbides would have exactly the same composition as fcc, and there would not be any chemical driving force. The only force then is curvature which makes them disappear like a very small grain in grain growth.
Please check in the .log file, the .TabLin file, or with the phase concentration outputs whether this is really the case. If yes, you could use a lower initialisation temperature (i.e. start checking for nucleation at lower temperature), use "limits" to prevent switching to the wrong composition set, or redefine the major compositions when creating the .GES5 file in case they have not been properly chosen (perhaps you also should just use FCC_A1#3 instead of FCC_A1#2...).
Bernd