Actually that's not technically accurate. Non-clustered indexes will have all the key columns of the clustered index, but that does not have to be the PK. Indeed, best overall performance often demands that it not be. For example, you might cluster by ( datetime, location, ident ) or by ( location, datetime, ident ) but have ident alone as the PK.
SQL DBA,SQL Server MVP(07, 08, 09) A socialist is someone who will give you the shirt off *someone else's* back.