Box-counting clustering analysis: Corrections for finite sample effects