>> Note: it is not necessary to have a FLAT flat for autocollimation, only a
SMOOTH one. 1/4 wave of curvature if it is smooth is VERY acceptable, which
you should be able to get easily by grinding three tools and mirrors against
one another.
For autocollimation SEVERAL waves of sphericity are acceptable. You can think of it as testing your mirror not by a starlight, but by a very distant pinhole.
But originally I think this flat was intended to be a coelostat. Not room for sphericity here at all, unfortunatelly (it would result in astigmatism).
Bratislav