For reasons of partial linguistic symmetry, say "quasi-factualism" for "quasi-realism." Now, suppose that there is an initial truth-value multiset [T, T] and a truth-value set {T, F} := {-1, 1}. These are both "initial objects," so to say, in some "category of truth objects." (Here, we have categories as subsuming both multisets and sets, depending on whatever appropriate parameters.)
Now, take the deminegative augmentation of the truth-value set to get intermediary truth objects t and f (with t := -i and f := i). Set factualism and fictionalism "across from each other" and have them be multiset-theoretic repeats of T and F, so that at this stage we have both the truth-value set {1, -1, i, -i} and the truth-value multiset [T, T, T, t, F, F, f] (assume that we preserve the original extra copy of T from earlier).
Quasi-factualism seems as if it is more often meant as an intermediary between realism and fictionalism alone, but would there be a quasi-fictionalist dual to quasi-realism, then? One might, for example, evaluate different domains-of-discourse/language games in terms of which subsets of the truth-value sets/multisets they are "faithful to," with the question opened as to whether there is any domain of discourse encompassing every possible such subset. Then the challenging declaration that moral information is not strictly a matter of realism can be altered into a range of problems: are there some moral claims that are instances of realism or some that are instances of fictionalism, quasi-fictionalism, or quasi-factualism as well? For would we need to keep detaining ourselves with saying that all sentences of this or that language game are only one of whichever class we have offered, here?
Justification of the initial truth-value multiset: note that a T-operator on propositions is such that ¬T = T¬, whereas with respect to the difference between absent and opposed functions like the two usual quantifiers and the two usual modal operators, ¬(¬X = X¬). Accordingly, the special lack of divergence for T by itself, here (modulo negation in general), is what fixes the multiset for us, whereas F is established from an elementary T = X case; but so we get both the multiset and the set "out of the deal."
Corollary: to be perhaps overly precise, the background explanation for the origin of the exotic truth values should actually provide for two more computed from the solutions to x + x = 1 and y + y = -1. I.e., there should be a half-true option and a half-false one even then, and further truth-values are computed as the solutions to x times x = 1 and x times x = -1 and then xx = 1 and xx = -1. In the 1-case, there, two "models" of arithmetic are varied over, one where 00 equals 1 and another where it's left as "indefinite." In the first case we get 0 directly as a special truth-value, neither true nor false directly as such, but which assigned in a special way to two sentences it then "computes to" a truth (or to no designated value at all, in the paramodel). Moreover, then, 10 = 1, as does (-1)0; but 01 = 0, so we have a pair of sentences such that when the second is true but the first is 0, the whole composite sentence is 0. Or we have the case 0-1, which = 1/0, so is also irresolute.