Re: vowel harmony

This reminds me of something I did in Okaikiar, that I had
a question about.

The singular/plural distinction in nouns is accomplished with
a vowel change.  It's not a simple feature change like ablaut
or umlaut; it's more arbitrary, although it could be described
as a rotation around the trapezoid:

        /&/ -> /e/
        /A/ -> /E/
        /E/ -> /i/
        /i/ -> /o/
        /o/ -> /A/
        /u/ -> /O/

Now the problem with this scheme in that it includes distinctions
based on vowel differences that are not phonemic in Okaikiar.
/&/ and /A/ are allophones, as are /o/ and /O/:  /A/ and /O/ both
appear before /r/; /A/ (but not /O/) also appears finally.

I accomplished this by including necessary environmental changes along
with the vowel change.  In the nominative case, singular
{zamo} [z&mo] becomes plural {zama} ['z&mA]; that's straightforward.
But in the accusative case, singular {zamod} ['z&mod] can't become
{zamad}, because that would be pronounced ['z&m&d] rather than
the ['z&mAd] required by the above list.  So instead it becomes
{zamard} ['z&mArd].  Going the other way, there are places where
a schwa is introduced after an /a/ or /o/ simply to separate it
from the following /r/; thus genitive singular {zama'} [z&'m&@]
is distinguished from nominative plural {zama} ['z&mA].

So my question: is this at all realistic?  I'm assuming at some
earlier point in the fictional history of the language, when the
grammar developed, the allophones were separate phonemes.  But
in real languages it would be more likely that the morphological
distinctions would disappear along with the phonemic ones,
so that the modern language would just have identical forms for
genitive singular and nominative plural in the third declension -
a rather Latinish feature.

Bear in mind that this language, while alien, developed among
humans.  They are Homo sapiens (although perhaps not Homo
sapiens sapiens), so I expect human linguistic universals to obtain.

Thoughts?

-Mark