As no answers have been forthcoming at this point, I will report what I have found: the issues I was struggling with in my question were covered by Paul Churchland, in his 1989 article, "Knowing Qualia: A Reply to Jackson".
In an earlier response to the Knowledge Argument, Churchland argued that it equivocates on the sense of 'knows about', but in doing so, he rephrased the argument in a manner that, Jackson said, missed the point. Jackson presented a succinct statement of the argument avoiding, he claimed, the misunderstandings of Churchland's version, but in "Knowing Qualia", Churchland asserts that this, too, is equivocal.
Jackson's concise statement of the argument is thus:
(1) Mary (before her release) knows everything physical there is to know about other people.
(2) Mary (before her release) does not know everything there is to know about other people (because she learns something about them on her release).
(3) There are truths about other people (and herself) which escape the physicalist story.
"Regimenting further, for clarity's sake, yields the following.
(1) (∀x)[(Hx & Px) -> Kmx]
(2) (∃x)[Hx & ~Kmx] (viz., "what it is like to see red")
∴ (3) (∃x)[Hx & ~Px]
Here m = Mary; Kyx = y knows about x; Hx = x is about persons; Px = x is about something physical in character; and x ranges over "knowables," generously construed so as not to beg any questions about whether they are propositional or otherwise in nature.
"Thus expressed, the argument is formally valid: the salient move is a modus tollens that applies the second conjunct of premise (2), '~Kmx', to the waiting consequent of premise (1), 'Kmx'. The questions now are whether the premises are jointly true, and whether the crucial notion 'Kmx' is univocal in both of its appearances. Here I am surprised that Jackson sees any progress at all with the above formulation, since I continue to see the same equivocation found in my earlier casting of his argument.
"Specifically, premise (1) is plausibly true, within Jackson's story about Mary's color-free upbringing, only on the interpretation of 'knows about' that casts the object of knowledge as something propositional... Premise (2), however, is plausibly true only on the interpretation of 'knows about' that casts the object of knowledge as something non-propositional, as something inarticulable, as something that is non–truth-valuable. What Mary is missing is some form of 'knowledge by acquaintance,' acquaintance with a sensory character, prototype, or universal, perhaps.
"Given this prima facie difference in the sense of 'knows about', or the kind of knowledge appearing in each premise, we are still looking at a prima facie case of an argument invalid by reason of equivocation on a critical term... The burden of articulating some specific and unitary sense of 'knows about', and of arguing that both premises are true under that interpretation of the epistemic operator, is an undischarged burden that still belongs to Jackson.
"It is also a heavy burden, since the resources of modern cognitive neurobiology already provide us with a plausible account of what the difference in the two kinds of knowledge amounts to, and of how it is possible to have the one kind without the other."
Jackson makes what he considers to be a crucial distinction between Mary coming to know what it is like for herself to see color, and knowing what it is like for the rest of us: while there was nothing to know about the former until her release (as Mary did not, in fact, know what it is like until she was released), we knew what it was like all along. Churchland's analysis shows that this does not avoid the charge of equivocation, which is perhaps not surprising (at least in my view), given that our knowledge of what it is like for other people (to the extent that we do know this) is just an assumption that it is, or closely resembles, what it is like for ourselves, and so depends on us knowing the latter.
Churchland goes on to point out that the Ability Hypothesis replies of Lewis et. al. also turn on this equivocation, by suggesting that what Mary learns on release is a form of 'knowing-how', which has long been recognized as being distinct from propositional knowledge. He says, however, that our knowledge of neurobiological processes means that we need not explicate Mary's missing knowledge solely in terms of her missing one or more abilities. He launches into a broad outline of what was then known of the biological processes of learning from sensory experience, extrapolating it into what he describes as "a live candidate for the correct story of sensory coding and sensory recognition that is not remotely propositional or discursive", pointing out that we share this biology with other trichromat animals that can make reliable color discriminations, but are lacking any linguistic ability. He also points out that Mary's lack of this sort of knowledge is not inconsistent with her "having stored, elsewhere in her brain, a detailed and even exhaustive set of discursive, propositional, truth-valuable representations of what goes on in people's brains during the experience of color. She may even be able to explain her own representational deficit in complete neurophysical detail" (Personally, I assume that Jackson would have said, and a physicalist could agree, that, in the circumstances of the first premise, she must necessarily be able to do so, and furthermore, be able to predict what changes will occur in her brain when she first sees something colored, so long as she can deduce how its image will fall on her retina. The ability to predict such changes does not, of course, imply the ability to bring them about by mental effort alone.)
Whether or not Churchland's outline of a scientific theory of color perception is true, he says, it is at least a logical possibility, and accordingly, "what we have sketched here is a consistent but entirely physical model (i.e., a model in which Jackson's conclusion is false) in which both of Jackson's premises are true under the appropriate interpretation [by 'under the appropriate interpretation', I take it that he means according to each premise's particular concept of 'knows about'.] They can hardly entail a conclusion, then, that is inconsistent with physicalism. Their composibility, on purely physicalist assumptions, resides in the different character and the numerically different medium of representation at issue in each of the two premises. Jackson's argument, to re-file the charge, equivocates on 'knows about'."
Later in this article, and more thoroughly in a postscript printed in "There's Something About Mary", Churchland analyzes various ways that the knowledge argument might avoid equivocation, but concludes that they all fail, one way or another.
I would like to thank @MarkOxford for patiently helping me understand various issues of logic and epistemology.
 Churchland, P., 1989, "Knowing Qualia. A Reply to Jackson", in "A Neurocomputational Perspective: The Nature of Mind and the Structure of Science", Cambridge, MA: MIT, pp. 67–76.
 Churchland, P., 1985, "Reduction, Qualia, and the Direct Introspection of Brain States", Journal of Philosophy 82: 8–28
 Jackson, F., 1986, "What Mary Didn't Know", Journal of Philosophy 83: 291–295
 Churchland's footnote:
[Translation: (1) For all knowables x, if x is about humans and physical, then Mary knows x. (2) There is something x that is about humans but Mary does not know it. (3) Therefore, there is something x about humans that is not physical.]
 Ludlow, Peter; Nagasawa, Yujin; Stoljar, Daniel, eds., 2004, "There's Something about Mary: essays on phenomenal consciousness and Frank Jackson's knowledge argument", Cambridge: MIT Press. ISBN 0-262-12272-3.