Our blogging site is going to go dark in an hour or so for a software upgrade, so I won't be able to post my full findings tonight.
What I can say is that, based on the current "rules of engagement" (i.e., the timer starts when the tables are populated), I believe I've got the execution time of the "discredited" syntax down to about fifteen seconds. Not bad for something that was originally going to take a couple of days to finish.
Why did I do this, when we've already got syntax that's clearly more efficient? Because I wanted to see what I was doing wrong that was making the discredited code perform so horribly.
I think I'm onto that now. Perhaps "Love Your Mistakes" should be Pond's Twelfth Law (although strict constructionists might argue that this exercise has been an application of Pond's Sixth Law, "Play With Your Product")..
I'll post more tomorrow when the site is back up.