It lowers the cost of the government performing wide surveillance of financial activity; the fear is that lowering this cost will enable more advanced forms of detection and enforcement that ultimately look and feel like harassment.
Currently, it would be more difficult to build this database as it's hodgepodged together from suspicious activity reports and subpoenas. There's of course the unknowable possibility that law enforcement and security agencies have secret ways of building a comprehensive realtime dataset - but, if they do exist, their secret nature reduces the scope of situations in which they can be used.