Lanczos vs AMS

Hello all,
I would like to ask that why we are able to solve natural frequency problem of a vehicle body with AMS solver whereas Lancoz solver gives the error of kinematic coupling* for the same problem.
Initially, while Lanczos solver solves the problem without any problem, I added a part and tied with the assembly and then I obtained such an error. (So there is no new coupling etc.)

*"THREE FACTORIZATIONS IN A ROW FAILED. CHECK THE MODEL. IT IS POSSIBLE THAT THE MODEL CONTAINS THE KINEMATIC COUPLING DEFINITION SET UP IN A WAY THAT A DEGREE OF FREEDOM HAS NEITHER MASS NOR STIFFNESS"


Best Regards