The "new" schematic may be a redesign by National.
Wolfgang's website has an old µa723 application note from Fairchild which shows a schematic very similar to your die.
An even more accurate match is the schematic from Motorola datasheet, but there are a few dumb and confusing errors. The circuit wouldn't ever work as drawn
Old TI datasheet (before National acquisition) also shows the same schematic as Motorola, without errors.
So there is quite a bit of agreement, and only National sticks out as the weird one. Some datasheets (ST, Philips) have no schematic at all.
And since Texas Instruments absolutely has no taste and looking at their schematics makes me want to throw up, I post corrected Motorola schematic below, which appears to match your die. Enjoy
edit
Added the Fairchild schematic for completeness. The only difference I can see is Q9 and Q11 collectors. I have no idea if the original Fairchild die matched the Fairchild schematic or maybe was like everybody else's.
I also don't understand what was the point of connecting Q9 and Q11 collectors to Q4 and Q5. VCC is available in the isolation well right underneath Q4 and Q5, they could have used that if they wanted to.