Sample data would be good as well

Using the following sample data

insert into Indexing values (199),(200),(210),(220),(230),(240),(250),(251)

insert into Indexing2 (Col4) values (199),(200),(210),(220),(230),(240),(250),(251)

I get no duplicates and now stream aggregate, the only difference is a left semi join on the exists vs a inner join.

Also post your execution plans in a SQLPLAN format so we can take a look.