So, I went back to Merlin Blencowe's site and re-read (for about the 5th time) his example on a SE design (his, with an EL34 power tube). The part about the screen resistor now makes a lot more sense (though I still could not explain it to anyone). The thing I missed or didn't understand well enough the first 4 times, was that the location on the load line graph of grid voltage curves is dynamic with respect to screen voltage/current, which changes as the current flow shifts away from the plate to the screen, as in near-0 and positive grid voltage conditions. So, the trick is to calculate the screen resistor size ONLY based on a 0-volt condition, when the screen is flowing most of the tube's current... BUT, it's not doing that most of the time, so for clean running (the dominant condition) you still calculate bias and quiescent operating point based on zero signal (which means about -20 volts on the grid in my case).
His three paragraphs on the EF86 pentode here are a good supplement to the SE example page with regard to screen current voltage, and a little more definitive, though pentode-general and not power amp specific (he also mentions by-passing the screen resistor with a capacitor, which I assume would work more or less like it does on a cathode to eliminate negative feedback and compression, but that's going to have to be another tutorial!).
I'll make a final determination on the exact value of the screen resistors once I get the real power supply up and running, but it'll be between 470R and 1k for these... just like everyone else does it (go figure) :-)