Weak vowel merger

Ie. are "roses" and "Rosa's" pronounced the same

For me those two words are clearly different, but 'rabbit' and 'abbot' rhyme. Also, "roses" is the same for me as "rose is", so I think my distinction between may be due to secondary stress on the final syllable of those words, rather than the lack of a weak vowel merger.
I've read that the Southern US generally preserves the distinction. Is it merged everywhere else in the US?

For me I do not have "roses" and "Rosa's" as homophones:

"roses" : ["Ro:zI:s]
"Rosa's" : ["Ro:z@:s]

but I do have "rabbit" and "abbot" rhyming:

"rabbit" : ["RE{:bI?]
"abbot" : ["E{:bI?]

Normally, I would say that I have the weak vowel merger except for in cases where /@/ is followed by a morpheme boundary, which do not necessarily merge with other cases (say, a morpheme boundary coming before /@/ or /I/ in the same position).

As I mentioned on another thread, I have a partial weak vowel merger. (My speech is different from Travis' though, because [@], rather than [I], tends to be my "default" weak vowel.)

I preserve [I] in word-final [Iz], [Id]:

"roses" [r\7UzIz] - cf "Rosa's" ["r\7Uz@z]
"stupid" ["stUupId]
"blessed" ["blEsId]

But elsewhere I generally have [@]:

"deliver" [d@"lIv@`]
"compatible" [k_h@m"p_h{4@b5=]
"rabbit" ["r\{b@t] - cf "rabid" [r\{bId]
"abbot" ["{b@t]

homophones for me:

"roses" : ["r\@}z@s]
"Rosa's" : ["r\@}z@s]

& "rabbit" and "abbot" rhyme:

"rabbit" : ["r\{b@?]
"abbot" : ["{b@?]

or in more careful speech:

"rabbit" : ["r\{b@t]
"abbot" : ["{b@t]

I'm somewhat confused about the application of the merger for me. As I said, 'rabbit' and 'abbot' rhyme for me, yet this is one of those things where I sometimes feel like they shouldn't, or that the rhyme is weaker than some others. Yet 'deliver' is [dIlIv@`], and 'compatible' is [km=p{4IbUu], with a clearly different vowel than schwa. I was reading the other day about secondary stress being a characteristic of Canadian English, and I think this maybe an example of it.
I think this difference with the merger is due to syllables with secondary stress not being fully reduced in Canadian English, and also that CanE tends to have secondary stress on the syllable before the primary stress (eg. [%dI."lI.v@`]. I still haven't figured out why 'compatible' should have [I] though.
What vowel does everyone have in the first syllable of 'exaggerate', 'except' and 'accept'? For me they are:
[%Eg."z{.d_Z3`.e?]
[Ek."sEp] or [Ek."sEpt]
[@k."sEp] or [Ek."sEpt]

For me, they're:

exaggerate [Ig"z{dZ@%r\eIt]
except [Ik"sEpt]
accept [@k"sEpt] or [Ik"sEpt]

I also have [I] in:

respect [r\I"spEkt]
esteem [I"stim]

I think I preserve [I] before nasals:

innumerable [I"nUum@`@b5=]
emaciate [I"meISi%eIt]
enough [I"nVf]

But not before [l]:

eleven [@"lEvn=]

And for the sake of comprehensivity, I should note that final [Ik] is another situation where I preserve [I]:

public ["p_hVblIk]
music ["mjUuzIk]
democratic [%dEm@"k_hr\{4Ik] - cf "haddock" ["h{4@k]

And, I suppose, also in final [Ig], like in "Zelig" ["zElIg]. And of course, in final "-ing" [IN].

On second thought, I think "enough" can be either [I"nVf] or [@"nVf] for me.

I suspected most other accents would have [I] or [@] for 'exaggerate'. That's another example of secondary stress preserving an unreduced vowel for me. However, I'm going to change what I said for 'except' and 'accept' because I realize now that those were citation forms. In regular speech, I produce [@k.sEp] or [@k.sEpt] for both. I tend to say 'enough' as [i."nVf] or sometimes [I."nVf].

In the case of word-initial weak vowels and weak vowels in unstressed initial syllables, I generally have free variation between [I], [1], and [@] in most cases, but generally use [I] or [1] before /n/ (as in "enough" [1~:"nVf] and [@] before /w/ (as in "beware" [b@:"we:R]), /l/ (as in "aloud" [@:"L\a:Ud]), and /r/ (as in "erect" [@:"REkt]).

However, though, many prefixes that historically had /I/ in them have been shifted so that they have a stressed form realized with [i] alongside an unstressed form with a weak vowel, as in "beware" also being pronounced [bi:"we:R] when stressed.

>>And for the sake of comprehensivity, I should note that final [Ik] is another situation where I preserve [I]:

public ["p_hVblIk]
music ["mjUuzIk]
democratic [%dEm@"k_hr\{4Ik] - cf "haddock" ["h{4@k]

And, I suppose, also in final [Ig], like in "Zelig" ["zElIg]. And of course, in final "-ing" [IN].<<

Actually, I have [I] consistently in these places as well, and do not have free variation there. That is, "haddock" aside, as I have ["hE{:4ak] for it.

Yeah, I have [I] in all those places you mentioned, Travis. I can hardly imagine what -ing would sound like with [@]. I also tend to use [i] in prefixes, as you mentioned, so I have enough [i"nVf] and erect [i"r\Ek]. Another example of my pre-stress non-reduction is 'democratic' often as [%dEmo"kr\{4Ik], though [@] does occur for the second syllable.