I simmed the basics of that 3100 Mk11 network ( I left out the 3 notch-filters ) .
The values for R3a + R3b resistors ( 24R & 62R ) do ( actually ) make sense since they end up being paralleled with the ( always in circuit 12R ) R3_ resistor.
It's a padding trick used within a few other higher-end networks from JBL .
The way the resistors add together allows the horns output in the 750hz ( crossover ) area to be virtually untouched ( since what gets attenuated is mostly over 1K ).
FWIW, placing an Lpad right by the driver will certainly attenuate any desired amount >> the problem is, it'll also attenuate the crossover region which wasn't part of the original network design ( from what I see ).
OTOH, increasing R3_ from 12R to 15R ( while holding everything else the same ) will give another 1.2db of mid-band horn attenuation ( while still keeping the crossover region mostly untouched ).
I have to believe the initial desire for 3 to 6 db more attenuation is a bit over-stated.