A constrains translation [x,y] and rotation [u,v] and B constrains translation [z] and rotation [w]. All 6 DOF are constrained, think about it physically as DOF constraint is driven by the physical realities of the geometry involved. If you put the rod inside a hole that simulates A and then up against an angled surface which simulates B, will you have any free DOF? How well/reliably B actually constrains rotation is another matter, which is affected by the diameter and the angle of the face (ie: a very shallow angle or a very small diameter will not constrain it reliably) however for your size rod at a 45deg angle it might work reasonably well.

Also small note, datum A is not defined as the center axis (ie: thats not what the datum feature symbol is pointing at, which defines your datum feature) - it is derived from the OD datum feature A (or technically its inverse datum feature simulator). We can talk about the theoretical datums, but its good to always remember the physical geometry from which it is derived. It may seem like I'm nitpicking, but its an important distinction.

Thanks chez311 that explains it very well. As you mention I think the caveat is if the surface is at a relatively small angle it might not do a good enough job of constraining the rotation, depending on your requirements.

If precision in the inclined feature relative to the hole is required then the hole would make a better reference to hold the part while the incline is made. One difficulty in evaluating such things is that analyzing the practical effect of such variations have on the operation in use. It might be a concentrated stress or a flow control that is the limiting factor over simple geometry.