OK, I measured the dual P73 module more carefully and the beam that exits the projector, measured at the scanner position, is actually 6mm wide by 3.2mm high. At a baseline of 17.7M, the spatially filtered beam is 14mm by 4mm. The beam should spread to 6/5 x 14mm or 16.8mm if it had been truly 5 mm in width. That calculates to 0.67mrad in the worst axis.
I discovered an interesting phenomenon however. The spatial filter uses a 1 power telescope with two symmetrical 75mm PCX lenses. This should not effect the near field beam dimensions and it does not. This should also not effect the far field divergence, but it does. Maybe the long FL lenses are easier to position to remove residual defocus out of the diodes that requires too fine an adjustment of the 4mm FL Optima collimators. Maybe the spherical lens pair introduces some spherical aberration that counteracts some over correction in the aspherical collimators. I actually think it might be some of both because it cleans up the beam as well and this is before the spatial filter blades are moved into position.