On a non‐uniqueness in fragmentation models