As far as forward throw, the plate is a lambertian emitter. This means ~80% of the light is in a 60 degree beam angle and nearly >95% within 70 degrees.
But of course, conservation of optical étendue ('optical entropy' so to say) can't be cheated and absorption and scattering losses will be the trade-off for better collimation.
Etendue shouldn't be too much of a concern because we are emitting onto a relatively large space. We hadn't considered using a BEF but will certainly look into it.
Someone else mentioned using a holographic printed film for the same purpose of a BEF. Have you every used something like this?
But BEFs are dirt cheap and do a very good job for this, your run-of the mill backlighting setup for screens is a very simple, effective system. Also: Using two linear BEFs (rotated relative to each other at a right angle) is more efficient than one pyramidal film.
The losses I mentioned are because of the light-recycling: You get lots more back reflections into the system - every ray that doesn't have the right exit angle gets reflected/refracted back into the system, the phosphor will absorb/re-emit/scatter, randomizing the ray's direction again and the ray 'retries' to get out of the system with a different angle.
These processes naturally induce losses (non-radiative decay in the phosphor, Fresnel losses at one of the BEFs, no perfect reflection on the back mirror). Even if you only lose a fraction of a percent, you lose that fraction lots of times before a ray may make it out of the system. It adds up - still, the collimation is very good for such a simple system (i.e. slap two films for a few dollars per sqm over your emitting area) it should be very easy to test.
Edit: You'll get the best collimation for 90°-angle prism films - there are other BEF types that'll smooth the angular distribution of the light recycling for wide viewing-angle displays.
If, however, blue LEDs and red LED emitted at similar efficiencies our technology wouldn't work. We'd lose too much energy in the conversion. Luckily (for us) blue LED efficiency is projected to increase more quickly over the next few years than other colors (more R&D dollars behind them).