Discrepancies between dbSNP Summary Data and FTP Data

The summary page of a particular build implies that there are7,231,721 rs numbers. I interpret your FTP files to have 7,238,948, a difference of 7227. Can you resolve this difference?

There are about 7,200 redundant rs numbers in the FTP file, because SNPs in the pseudo-autosomal region (PAR) are reproduced in both the X and Y chromosome files. The introduction of PAR into the NCBI assembly is fairly recent. The rs set in our FTP files used to be unique—not so any more.