Thanks. I had the impression that one should use execution plan to get an idea of how well it'll perform for different applications whereas timing the result would be too specific to only one particular data set/schema.

Also, now that it's obvious that strawberry query isn't the best choice, I wonder what else good a triangular join would be for? Jeff Moden had cited them as one of "hidden RBARs" but also mentioned in pass that they can be useful in rare instances. Does anyone know of such instances?

Thanks again!