The origin of the transition of the flow structure between free shock separation and restricted shock separation in a compressed truncated perfect nozzle is studied. A three-dimensional compressible Navier-Stokes solver is used to capture the nonaxisymmetric unsteady flow structure inside the nozzle. The transition is numerically reproduced under the experimental condition where the transition occurs. Until now, the transition has been numerically observed in a thrust optimized nozzle, a parabolic nozzle, and a highly compressed truncated perfect nozzle, where an internal shock generated from the nozzle throat is the characteristic. However, in the compressed truncated perfect nozzle studied here the internal shock is weak and cannot be detected by the well-known shock function. Therefore, a function is used to detect the compression and expansion and it is found that a low-expansion region exists inside the compressed truncated perfect nozzle. The transition is likely to occur when this region is broadened radially at a mixture ratio lower than the design value.