bunched/molar r

When I was still a schoolboy I found Daniel Jones’s Outline of English Phonetics in the local public library and devoured it with great interest and excitement.
It all made perfect sense to me, with one exception. That concerned the English r sound. I couldn’t square Daniel Jones’s description with what I could feel myself doing when I pronounced this consonant.
Obviously I was not using a “rolled” r (trill) or a “flapped” ɾ (tap). And I wasn’t making any kind of uvular sound. So I must be using the “most usual English r”, the “fricative lingual sound” ɹ (which we would today call a postalveolar approximant).
But I didn’t seem to be articulating it with the tip of my tongue so much as with the body. And the combinations kr (as in crown) and ɡr (as in green seemed to have some sort of affricated pronunciation.
My puzzlement was resolved only several years later when John Trim explained to me that I use a “molar r”. The strange thing is that it doesn’t sound any different from the postalveolar r sounds that other people use. People don’t react by saying “that’s a funny kind of r”.
The VASTA listserv has been having a discussion about “bunched r” recently (another name for the same thing), and I said I would write something about it here.
Here’s Catford, talking about the AmE NURSE vowel:

This vowel was formerly described as ‘retroflexed’ but this is not a correct description. It does not usually have the upward curling of the tongue that is characteristic of retrofexion. Instead, the main body of the tongue is bunched up into a kind of half-close-central position, but with two peculiar modifications: one modification is a moderate degree of deep pharyngalization: the root of the tongue
is drawn back into the pharynx just above the larynx. The second modification is a fairly deep depression in the surface of the tongue opposite the uvular zone. This sub-uvular concavity can be acquired as follows. Produce a uvular trill. Note that in order to do this you have to form a longitudinal furrow in the tongue within which the uvula vibrates. Now move the whole body of the tongue slightly forward, while retaining precisely that deeply furrowed configuration. The result should be a close approximation to the typical American ‘bird vowel’, for which the phonetic symbols [ɜ˞] and [ə˞] have been used — both representing a central vowel with an r-like modification.
As we saw, this very strange American vowel involves not only a concavity — or ‘sulcalization’ (from the Latin sulcus `a furrow, or trench') — of the tongue in the neighborhood of the uvula, but also some slight degree of pharyngalization. It is because of this that a series of vowel-sounds with modification of this rhotacized type in some languages spoken in the Caucasus area of Russia, notably Tsakhur and Udi, are known as ‘pharyngalized’ vowels.

Like Erik Singer, who found this quote from Catford, I'm not sure I agree with the claim of pharyngalization. It can be there, no doubt, but I don’t think it's a necessary accompaniment.
My own r is consonantal, of course, while the American vowel is syllabic (a vowel). But the articulation used appears to be the same.
In my experience, when someone has claimed that there is an audible difference between the molar and the postalveolar kinds, I find that there is also some kind of difference in secondary articulations (pharyngalization, labialization etc), but no audible difference when this is stripped out. That’s why I have some sympathy with the argument that if we can’t hear the difference we don’t need a special IPA symbol for the molar r. It will be sufficient to symbolize the secondary articulation(s), e.g. ɹ^ʕ or ɚˁ.
But the consensus on the VASTA list seems to be that we need a special symbol. Erik Singer said

I very much agree that we need a symbol for the “bunched,” “braced,” or “molar” /r/. If it has to go in the set of IPA extensions for disordered speech, so be it. At least we’ll have a symbol. I wonder, though, should there be separate symbols for the vowel and consonant? It’s really a single physical action (varying only, I think, in the degree of the bracing), and according to the rules of the IPA, there should therefore be only one symbol corresponding to it. […] There simply must be a unique symbol. This is my awkwardly bitmapped attempt to create a symbol that takes the turned-r and adds a another wing to it, stretching out to the right. We can recognize this as r-ish, but it is sufficiently different from the alveolar-approximant symbol to avoid confusion with that placement. This symbol could also be modified to specify degree of bracing, with one or two additional “wings” stretching to the right, at mid-level and at the top.

Amy Stoller says she associates this sound (a "hard-R" sound) particularly with an Oklahoma or Texas accent, though it is also found elsewhere. ‘Michael’ added

It is not impossible to make a "lighter" sound with this placement —though it becomes a bit more of a mid-tongue bunching rather than the hard retraction of Oklahoma.

Can people actually hear a difference based purely on place of articulation? Or is there always something else involved? Let’s have some sound files.