In case it helps, here's how I (/people around me) would deal with those words...
Possible vowel reductions:
abominable, accuracy, accompaniment, apothecary, cemetery, interesting, laboratory, literature, miserable, oratory, solitary,
These reductions form a continuum. At one end is "interesting", where I think NOT reducing the vowel would probably indicate some sort of ironic overpronunciation. "Literature" follows close behind, along with "solitary"; I also learned "cemetary" without the reduced vowel, and while I tend to include it now, it's effectively an affectation for me. "Oratory" is a similar case: I think I by default have it without the vowel in the sense of 'rhetoric', but I have a version with the vowel in the specific meaning "Catholic Order following the teachings of St Philip Neri, or place of Christian worship associated with that Order". "Laboratory" is an unusual one for me - my 'native' pronunciation has initial stress, with occasional reduction of the following schwa, probably less than expected due to the taboo of "lavratory" nearby; however, these days I mostly stress the second syllable, in which case the penultimate vowel is usually but not always reduced.
Near the other end of the continuum, "abominable", "accompaniment", "miserable" and "voluptiousness" are all words I would keep the vowel in in careful speech, and that I honestly think I at least sometimes have the vowel in in even rapid speech - but where I'm sure I do drop the vowel sometimes. "Apothecary" is more often reduced than these, but less often reduced than the previous category of words.
And then AT the other end of the continuum, there's "accuracy", where dropping the vowel is disallowed for me, but where I think it happens sometimes anyway. The additional confusion is that of course I don't have coda /r/, so this is a 'vowel reduction' from /Ur@/ to /U@/, and may be considered a consonant reduction instead. This sounds like 'accuecy', which is why it's disallowed - but as I say, I think it happens sometimes in rapid speech, possibly with some residual subphonemic rhotic element.
I've also heard 'testimony' with the penultimate vowel dropped, although I don't think I do that myself generally.
As can be seen, these tend to be cases in which an unstressed vowel is followed by a sonorant.
Polyphthongs
cauliflower, voluptuousness
These have antepenultimate stress phonemically, because -ower and -uous have phonemic triphthongs, and form single syllables. Phonetically, in over-annunciated speech, these triphthongs (aU@ and juU) may be broken up with epenthetic /w/ (aUw@ and juwU), but this is both phonemically and etymologically secondary, and don't seem to effect stress patterns and phonotactics and the like. [in much the same way that, in over-annunciated speech, "goal" may be said to have two syllables, but generally doesn't in actual fluent speech].
Coincidentally, these triphthongs in careful speech are often reduced in rapid speech anyway (aU(w)@ > a:, and tju(w)U > tSU).
Secondary stress
agglutinating, agriculture, architecture, caricature, dedicated, dictatorship, discriminating, doppelgänger, egalitarianism, isolating, invigorating, lesbianism, orthodoxy, patriarchy, witticism,womanizer
The suffixes -ate, -arch and -dox all seem to generally take secondary stress in these words. Notably, -ate is a suffix with phonemic secondary stress for me - secondary stress in verbal forms, and in some nouns, and no stress in adjectives and most nouns. [of course this could also be analysed as a vowel quality and length distinction, but I think it makes more sense to view stress as primary and reduction as secondary].
More difficult are the suffixes -ism and -ship... these vowels should only reduce to /I/, rather than /@/, and it's possible that this is leading me to ascribe secondary stress when there isn't any. I'm pretty confident there's secondary stress in 'lesbianism' and probably 'egalitarianism', I'm a little less sure of 'witticism', and I'm torn on 'dictatorship', though I think it does have the stress.
"Architecture" and "agriculture" likewise have secondary stress on the penult, despite lacking one of these affixes (and it becomes primary stress if you add another affix (agricúltural, etc), as does "doppelgänger". "Caricature", on the other hand, defies this pattern by having secondary stress on the final, NOT on the penult.
Full dual stress compounds
brother-in-law, candle-maker, counter-tenor, metal-worker, money-lender, trust-worthiness
These words are pronounced as compounds, even when not spelled as such. The second stress is usually weaker than the first, but stronger than routine secondary stress. In all cases other than countertenor, the primary stress can be moved for emphasis, although this isn't obligatory. ["It's terrible how many candles people use!" - "not if you're a candle
maker!"]
Different stress locations
oligarchy, partisanship
In the case of "partisanship", I generally have penultimate stress; this matches the adjective, and the name of the Yugoslav Partisans; I can also sometimes say it with initial stress but strong secondary stress on the penult, to match the noun when not used for the Yugoslavs. I also have a pronunciation with initial stress and no secondary penultimate stress, used only in the context of political science, and most often in discussing American politics, which is clearly a loanword in this context - this is like 'dictatorship', in that I think there's a secondary stress on the final, but I'm not certain.
In the case of 'oligarchy', second-syllable stress is normal for me. However, I can optionally have initial stress, with secondary stress on the penult, when talking specifically about a political system dominated by oligarchs (rather than, say, in economics).
Residual words with genuine, unambiguous, single, preantepenultimate stress:
adequacy, gentlemanly, intimacy, inimitable, mercilessness, nominative, seriousness, veritable
The odd one out here is 'nominative'. This 'should' have secondary stress on -ate, in accordance with the general rule about. Indeed, it does - in any context
other than linguistics. For me, all case names in -ate lack secondary stress, even when they would have it in any other context. [for instance, when discussion physical ablation, 'ablative' has at least secondary, and possibly primary stress on the second syllable, but the linguistic case, for me, does not].
