Distinguishing effects of area per se and isolation from the sample-area effect for true islands and habitat fragments