The fact that fewer rows are processed doesn't necessarily mean the query runs faster. Note that "explain" shows that the "in" form requires two sorts and a hash join, while the "join" form requires one sort and a merge join. hash joins can be expensive as the hash tables have to be built before the join can commence. Then again merge joins require the operands be sorted (which is why one is being explicitly sorted).
Both of these plans look like good ones to me. As to which runs faster you either need to find a slower computer (since they both report "CPU 0.00") or insert a lot more data ...
