You are very thorough and technically correct in your response. As you have a thorough understandng of receptor activity you will no doubt recognize that what I have posted is also technically correct but dumbed down just a bit for the average-advanced member.
Regardless of which specie is providing the density the basic requirements will be a partially positive hydrogen attracting an electron pair (at its most basic level). As you point out it can be a keto group, nitrogen, or oxygen atom that can act as the donor.
What we have is a basic question of which is more influential in determining receptor affinity: bonding potential at C-3 or conformational structure? IMO, the electron delocalization (hence stability and also structural integrity) afforded by the extended conjugation of the 4,9,11 or even just the 9,11-ene's are more influential for receptor affinities. It is true that a typical 3-keto has enough bonding potential to affect conformational distrortions of the entire ring structure when C-17 is viewed as a weak attachment point for planar alignment. In a nutshell, if the need to strongly attach at C-3 is bypassed via diene or triene induced structural rigidity then the importance of that hydrogen bond is decreased or perhaps even eliminated. This argument applies to 4,9,11 triene's or 9,11 diene's in general.
With respect to the molecule that you presented, the notable absence of the 4,5-double bond will allow the alpha ring to assume a full chair conformation. Thus, even as a 3-ol the molecule should present a high affinity due to steric location and decreased proximity. In theory, the lack of alpha ring unsaturation should allow affinities that are in magnitude with 3-alpha's.
...which leads me to my reply to your diol comment: a 3a,17b-diol is VERY active compound (more so androgenic than anabolic but both values are increased dramatically). This is basis for a simple observation that even 3-ol's can bind to receptors strongly if given the proper steriochemistry...such as with an unsaturated alpha ring substituted 3b-ol.
As for why I believe your values are off: all my references indicate RBA's for a general 4,9,11 triene in the magnitude of 500 each (V.P., S.V., and L.A.). This is regardless of substituents at C-17...acetates up to undec's. I have not referenced the base molecule values so I could be wrong but the trend of increased activity with conformational rigidity cannot be denied.
Chemo