I think the confusion may stem from the fact that the Broad Institute's GSEA tool uses a specific way of testing for enrichment of entire classes/sets of genes, which is not ORA, but "functional class scoring" in the terms of the review you're citing. The main difference between the methods is captured by neither nomenclature, IMO. The techniques that Khatri et al refer to as "ORA" basically just compare the absence/presence of single genes or gene sets between two lists (think Venn diagrams) whereas the FCS techniques refer to methods that try to take more information into account, such as the ranking of one of the lists.

The ranking is not determined by FCS. FCS needs a ranked list of your genes of interest to determine whether the genes that are at the top/bottom of your list are more likely to be present in the gene sets that are to be assessed for overrepresentation.

This question is answered, for that reason I close the question.