Are fault growth and linkage models consistent with power-law distributions of fault lengths?