In my opinion, A&G make too big a fuss over mixed i-stems, which are essentially a broad class of nouns that usually feature a genitive plural in -ium and might feature an accusative plural in -īs, but are otherwise regular third declension nouns.
Here are five of their six morphological classes, along with all the examples they offer. (I’m excluding 71.4 because it’s ridiculous that they’ve placed it there at all):
1. Nouns in (n.) -ēs, (g.) -is.
- acīnacēs, -is (m.) scimitar
- aedēs, -is (f.) temple
- aciēs, -is (f.) point, battle line
- caedēs, -is (f.) slaughter
- cautēs, -is (f.) crag
- clādēs, -is (f.) destruction
- compāgēs, -is (f.) structure
- contāgēs, -is (f.) sense of touch
- famēs, -is (f.) hunger
- fēlēs, -is (f.) cat
- fidēs, -is (f.) faith, trust, loyalty, reputation, etc.
- indolēs, indolis (f.) inborn quality
- lābēs, labis (f.) fall, destruction
- luēs, luis (f.) liquid water
- mēlēs, mēlis (m/f.) marten, badger
- mōlēs, mōlis (f.) mass, bulk
- nūbēs, nūbis (m/f.) cloud
- palumbēs, palumbis (m/f.) dove
- prōlēs, prōlis (f.) shoot, offspring, descendants
- prōpāgēs, prōpāgis (f.) shoot, offspring, descendants
- pūbēs, pūbis (f.) young man
- sēdēs, sēdis (f.) seat, office
- saepēs, saepis (f.) hedge, fence
- sordēs, sordis (f.) filth
- strāges, strāgis (f.) overthrow, destruction
- struēs, struis (f.) pile
- subolēs, subolis (f.) shoot, offspring, descendants
- tābēs, tabis (f.) decline, decay
- torquēs, torquis (f.) necklace
- tudēs, tudis (m.) hammer
- vātēs, vātis (m/f.) prophet
- vehēs, vehis (f.) cart-load (quantity)
- veprēs, vepris (m.) bramble-bush
- verrēs, verris (m.) boar
- vulpēs, vulpis (f.) fox
2. All monosyllable nominatives in -s or -x preceded by a consonant.
- ars, artis (f.) skill, art, technique
- pōns, pontis (m.) bridge
- arx, arcis (f.) fortress
3. On the following monosyllable nominatives in -s or –x preceded by a vowel.
- dōs, dōtis (f.) dowry
- fraus, fraudis (f.) deceit, fraud
- glīs, glīris (m.) dormouse
- līs, lītis (f.) case, quarrel
- mās, māris (m.) male
- mūs, mūris (m/f.) mouse
- nix, nivis (f.) snow
- nox, noctis (f.) night
- strix, strigis (f.) channel, furrow
- vīs, vis (f.) force
4. Polysyllable nominatives in -ns or -rs.
- cliēns, -entis (m.) client, follower
- cohors, -ortis (m.) companion
This does not apply to all present active participles!
5. Patrials (nouns denoting birth, class, abode) in -ās and -īs.
- Arpīnās (Arpīnātēs)… Aprīnātium
- Optimās (Optimātēs)… Optimātium
- Penās (Penātēs)… Penātium
- Quirīs (Quiṝitēs)… Quiritium
The (very much non-)Essential AG: 71-2
Latin for Addicts 
Hello, rsmease!
Tell me please, if you have an e-mail that I could write to, coz, translating the texts I’ve got so many questions constatly arising, that I feel ashamed to leave them all here, not correspondent to the lessons-messages in your blog. So I thought we might chat more via email.
Thank you for the information!
How about Facebook, instead?
https://www.facebook.com/rmease
I can respond to your questions through the messaging system.
The page you’ve given is “temporarily unavailable to me” – moreover, I’m not registered at facebook, so I’m afraid this is not the right way for me… sorry… do you think e-mail system might still be appropriate?
rmease@uchicago.edu
There seems to be so much confusion among these stems, do you suppose the actual Latin spoken, not written, would be less confused? So many nouns used in different declensions or subtypes of declensions among different speakers, different suffixes too at times. If Latin were alive today they probably would have standardized and made one dialect’s variations the standard and possibly we would have less irregularities.
I would like to think that, but if you look at the chaos of the English language, I’m doubtful that a modern Roman empire would have bothered to fix these kinds of flaws in their morphological system. If English can do just fine with though/through/thou, the Romans would have managed to make crazy Latin i-stems work as well.