apostrophes in transliteration (stress in Tokana)

I'm considering adding apostrophes to Tokana to separate clitic pronouns from their
hosts.  I recently made some changes to the morphology of Tokana which in turn led
to a radical overhaul of the stress system.

The old stress rule was as follows:

(1) Words ending in a consonant or diphthong stress the final syllable--e.g.
_totsat_ "table, bench" [to.TSAT], _ikei_ "dog" [i.KEJ].
(2)  Words ending in a non-diphthong vowel stress the penultimate syllable--e.g.
_paua_ "wash" [PA.wa], _itskana_ "arrive" [its.KA.na].

Exceptions to this rule were confined to a handful of words ending in a
non-dipthong vowel with final stress, which were all marked with an accent grave,
much as in Italian--e.g., _nape'_ "daughter", pronounced [na.PE].  The final stress
here is justified by the fact that these words used to end in an [h]; the [h] was
dropped, but the word-final stress triggered by the final consonant was retained.
(In fact, the [h] remains when a suffix is added, e.g. _napeh-mo_ "with the
daughter".)

In my original conception, subject & object pronouns suffixed to verbs, as well as
possessive pronouns suffixed to nouns, triggered a stress shift:

  soiha    "wife"      [SOJ.ha]
  soihama  "my wife"   [soj.HA.ma]

But now--for reasons which are too convoluted to go into--I have decided to change
the rule so that suffixed pronouns no longer trigger a stress shift:

  soiha    "wife"      [SOJ.ha]
  soihama  "my wife"   [SOJ.ha.ma]

As a result, I now have oodles more exceptions to the basic stress rule than I had
before, and the question is, how to mark them?  This is especially important
because of some other morphological changes I've made:  Originally I marked the
difference between main clauses ("I go") and embedded clauses ("...that/if/when I
go") by adding a suffix to the verb:

  etum     "X walked"                 [e.TUM]
  etuma    "that/if/when X walked"    [e.TU.ma]

This is still true in most cases.  However, there are now a couple tense forms
where the difference between main/embedded clauses is marked solely by a shift in
stress (the story is that historically there was a suffix, but it got reduced).  As
with the irregular forms mentioned above, these forms are marked with an accent
over the final vowel.

  eta              "X walks"                [E.ta]
  eta' (< eta-a)   "that/if/when X walks"   [e.TA]

As a result of this change, there are now lots of minimal stress pairs in Tokana
(words which differ only in which syllable gets the stress).  Now this is all fine
as far as I'm concerned, but the question is, how should I indicate the stress in
cases where a non-stress-shifting pronoun is suffixed to one of the forms above?
For example, suppose that I add the suffix "-na" (he, she) to the forms above:  How
do I distinguish orthographically between [E.ta.na] "s/he walks" and [e.TA.na]
"that/if/when s/he walks"?

I envision three possible solutions:

(1) Do like in Spanish: Words which obey the basic stress rule of the language are
unmarked, while words which disobey it are marked with an accent over the stressed
vowel.  The disadvantage of this is that it would lead to a proliferation of
accented vowels, which I find aesthetically displeasing.

(2) Mark off suffixed pronouns with an apostrophe, to show that they do not
participate in determining the stress of the word they attach to.  This would lead
to a proliferation of apostrophes, which I would like a bit better than a
proliferation of accent marks, but not much.

(3) Add an addendum to the basic stress rule of Tokana to the effect that suffixed
pronouns don't cause a stress shift, and then mark exceptions to THAT rule with an
accent over the stressed vowel.  Under this system, [E.ta.na] would be unmarked (it
obeys the rule which says to first strip off the suffixed pronoun, and then assign
stress to the penultimate syllable if the remainder of the word ends in a
non-diphthong vowel), while [e.TA.na] would get an accent mark over the stressed
vowel (it violates the above rule).

I'm leaning towards some sort of compromise between (2) and (3), but I'm not sure
yet.  Any suggestions or comments?

Matt.