Yes, Lubinski & Benbow did that research on the old SAT, so those estimate formulas do not apply to recent SATs.
As to the ACT scores for the Triple Nine Society: that is a private organization; I cannot speak with any certainty to why they continue to accept the ACT. I assume their psychometric consultants have some data that support this decision. My speculation would be that the fact that the ACT has not changed significantly in decades helps them to have some standard of comparison. It's also the case that achievement tests do have some correlation to intelligence (overachievement is not hypothetically possible, after all), and a composite score of 34 is a pretty low percentage occurrence. Based on the published tables, it appears the population mean in the standardization sample was a score of about 16, with a SD of about 6, and +3SD thus falling at 34, corresponding to the 99.9th %ile. Note that DYS also allows certain achievement tests to be used as proxies for cognitive ability.
The current SAT is less usable for this purpose, when given to typical-age students, because the 99th %ile is about 780, which doesn't leave much headroom to spread that last percentile. It still has some utility for above-grade-level testing, which is why the major talent searches still use it, but since they keep changing the test, it becomes more difficult to do longitudinal research and comparisons. (Also see the comments I made in my previous post.)
More straightforward method of estimating your rank in the population (not your IQ):
Here's the percentile table for the PSAT 8/9 the year you first took it:
http://www.nwgjb.com/upload/files/20161010155620550.pdfThe national 50th %ile for total score is 820, with a SD of about 155, which puts the 99.9th %ile at about 1285.
I hope this satisfies your curiosity, at least a little bit.
