This problem in relation to ordinary language is related to a similar issue in Russell's {\it Theory of Types}
as used by Whitehead and Russell in {\it Principia Mathematica}.
Russell uses the term "incomplete symbol" for those apparently denoting expressions which are in that system
explained in context without assigning a designatum to them.
The use of incomplete symbols for classes in {\it Principia Mathematica} yields what Russell calls a
"no class" theory.
In the context of Russell's theory of types this permits talk as if of extensional classes in an
underlying ontology of intensional propositional functions.
If similar devices are used in a pure set or class theory, they are less successful.
This is because it is necessary to quantify over classes, and for that reason there
must be something which the set designates.
In The Theory of Types, talk of classes reduces to talk of propositional functions,
and the necessary quantification can take place over the propositional functions.
The full story on exactly how much you can do in set theory with inconplete or syncategoremic
symbols appears in Quine's "Set theory and its logic" (1960), the development of the
implications for ontology in ordinary language begins here and appears in its mosts influential
form in "On What There Is".
The flaw in the ointment here in its intended application to ultimate ontological
questions is the presumption that we cannot quantify over things which do not exist,
which seems on its face to be contrary to reasonable usage.
How many cases did Sherlock Holmes solve?
Is there any reason why we cannot generalise about these cases, if they have
anything in common, would we not say that "all" the cases of Sherlock share
that characteristic?
The contemporary controversy while Quine was developing his position on
"ontological commitment" was with Carnap, whose explicit position as articulated
in "Empiricism Semantics and Ontology" is a denial that use of a language
involving quantification over some domain of entities (particularly abstract entities)
can be undertaken on a pragmatic basis without either "committing" to the ontology
in a metaphysical sense, or even aknowledging that the metaphysical question has
any meaning.
(Of course, Quine had an answer to Carnap.)
Quine's closing paragraph in which he takes the the range of variables to demonstrate
the limits of sincere nominalism, may possibly have some force against the nominalist,
but seems to me to have no force against the idea that ontology is conventional and
contextual, and the range of variables in such a context says nothing absolute
or in other contexts about what exists.