Cobra's drawing is dimensioned fully, unlike the drawing from post #1. Cobra has .050 of freedom per side at the ball pivot. The shank and tool have a 1:1 length to pivot ratio, so that the .050 of movement at the eccentric shank is converted to a proportional displacement as rocking motion at the tool. The pivot has enough range of motion to not bind. The tool tip tracks on center and the broaching motion cuts.
I don't think your tool is rocking enough, which is why it isn't cutting. Possible causes are not enough shank eccentricity, or tool motion binds somewhere and the loss is preventing rocking due to clearance or due to changing the proportionality of the tool lengths on either side of the pivot.
View attachment 454495
Clearances of mating parts not shown above.
View attachment 454494
Cobra's plans show all dimensions.