>Unofficially, in a table with 400,000 rows, both are taking the same amount of time. (less than half a second on an unindexed table, and then instantly when I create an index)

>I think both queries are good. I just prefer CTEs for things like this.

That does not backup your claim about single table read vs multiple table reads. I can't see your claim being substantiated. As for your test, time does not say me anything. What does the profiler say ??