In national research assessment exercises that take the peer review approach, research organizations are evaluated on the basis of a subset of their scientific production. The dimension of the subset varies from nation to nation but is typically set as a proportional function of the number of researchers employed at each research organization. However, scientific fertility varies from discipline to discipline, meaning that the representativeness of such a subset also varies according to discipline. The rankings resulting from the assessments could be quite sensitive to the size of the share of articles selected for evaluation. The current work examines this issue, developing empirical evidence of variations in ranking due changes in the dimension of the subset of products evaluated. The field of observation is represented by the scientific production from the hard sciences of the entire Italian university system, from 2001 to 2003.