Functions of Classifiers (in a conlang)

Well, for a project I'm working on I'm trying to develop a system of
noun classifiers, in the sense that the term is used in "Classifiers: A
Typology of Noun Categorization Devices" by A. Aikhenvald. She says:

" Noun classifiers characterize the noun and cooccur with in a noun
phrase. They have been recognised in Australian (Dixon 1977; 1982;
Wilkins 1989; Sands 1995: 269-70) and in Mesoamerican languages (Craig
1986b: 1986c). Their properties have been discussed by Craig (1992;
forthcoming; 3.2.3 below). In the Australianist tradition, they are
called 'generic classifiers' or 'generics' (Sands 1995: 269-70; Harvey
and Reid 1997: 9-10).
 The definitional properties of noun classifiers is that their presence
in a noun phrase is independent of other constituents inside or outside
it. Their 'scope' is a noun phrase. They are a type of non-agreeing noun
categorization device, their choice being determined by lexical
seletion, and not by matching any inflectional properties of nouns with
any other constituents of a noun phrase.
 Additional, contingent properties of noun classifiers are:

(i) The choice of a noun classifier is based on semantics. Every noun in
a language does not necessarily take a classifier. (3.2.1)

(ii) Languages may allow the cooccurance of several noun classifiers
within one noun phrase (3.2.1)

(iii) One noun can be used with different classifiers, with a change in
meaning (3.2.2)

(iv) The size of the inventory of noun classifiers can vary, from a
fairly small closed set to a fairly large open set. Consequently, noun
classifiers can be grammaticalized to varying extents (3.2.3)

(v) Noun classifiers are often used anaphorically; they may
grammaticalize as markers of syntactic functions (3.2.4)"

Aikhenvald contrasts noun classifiers with the numeral classifiers of
many East Asian languages, with noun classes (exhibit agreement,
normally bound morphemes), and with other kinds of classifying device.
She also mentions that in some languages classification of nouns using
such a marker is required ("Akatek (Kanjobal Mayan: Zavala 1993: 25-7
and p.c.) has 14 noun classifiers; they are obligatory as nominal
adjuncts, and can be used anaphorically").
 This was what I was planning to do... make the classifiers required
with all NPs, with differences in meaning corresponding to difference
choices of classifier. The problem comes when considering two possible
extensions in the function of such classifiers that, in the case of my
conlang at least, clash:

(i) As the quoted text from Aikhenvald mentions, many languages use
their noun classifiers as pronouns. This is a natural extention.

(ii) Another possibility is to use the classifiers as THE marker of
nominality, ie nominalizations of finite verbs can be carried out simply
by giving the verb a classifier, and the lack of a classifier can turn
any stem into a predicate (thus making nominality explicitly marked
rather than associated with certain stems).

Now, there is an issue with combining these two ideas: with the word
order rules I was considering, and I think probably with any possible
word ordering, it's too easy to find examples of sentences where the
classifier is ambiguous in whether it is acting as a pronoun or as a
nominalizer for a following verb. Eg:

MAN.CLS man come and MAN.CLS angry shout

Is this:

The/a man came and he shouted angrily

or...

The/a man came and the angry (man) shouted?

 Because of the problems with such a combination, I don't think it's
realistic to combine the two possibilities (unless someone can cite in
detail a natlang precedent). And I'm undecided which to go for... the
use of classifiers as pronouns is a pretty good way of reference
tracking and is certainly minimalist in that it does not require
additional third person pronouns, but making classifiers the true
markers of nominality and relativizers seems to me to be an elegant use
of the idea too.
 The issue if I go with (ii) is what to do about third person pronouns.
There seem to be two major possibilities:

(i) Have a separate unrelated set of third person pronouns. This would
mean losing distinctions though, since I have no desire to invent
unrelated pronouns for each noun classifier.

(ii) Somehow build pronouns on classifiers. For example, some ending
could be added to the classifiers (perhaps an erstwhile generic root?) Eg:

CLS.MAN-PRONOUN.MARKER = "he"

The second solution is easier, but has the downside of perhaps making
pronouns longer... and, despite the fact that some languages have (some)
long pronouns like Spanish "nosotros" or Japanese "watashi", I
personally prefer my pronouns, which are regularly used words, to be
short. Especially since those languages where pronouns are long (in
certain circumstances.. compare the Spanish object clitic nos with the
independent form nosotros) are typically those in which in many
circumstances pronouns can be dropped (Spanish is pro-drop with
subjects, and Japanese just loves dropping everything).
 Anyway, this has mainly been me thinking aloud, but I would welcome
comments and suggestions please?

Replies