You could be correct in your conclusion regarding the accuracy of the "unrestrained" displacement/rotation.

The displacements will not cancel the flexibilities, because the nozzle and pipe are connected with a CNODE. The Nozzle is displaced, but the flexibilities are between the nozzle and the pipe. The pipe node and the nozzle node may see different deflections based on the "stiffness-load" values.

Even modeling the tank (which I recommend) won't help with the bulging, since CAESAR II uses 3d Beam elements. There is no accounting for shell behavior.