Franx, you have done a fantastic job, including the discovery that if the unknown chr (‘) was turned into U+02BB things worked better. Now you have mentioned U+02BB my comments below deal with both your proposed solutions and the U+2018/U+02BB problem in Polynesian languages - maybe this should move to another forum, but I leave that up to you.
I should have realized that this import problem would drag me straight back into the handling of the glottal/hamsah in Polynesian languages. The glottal is used in Hawaiian (called ‘okino), Tahitian, Tongan, Samoan and Cook Islands Maori (what I am dealing with), but is not required in NZ Maori (a.k.a. Maori). By the way, I neither speak nor write CK Maori, but I do strive to record the CK Maori names for plants and animals in a modern orthography.
U+2018 (Left Single Quotation Mark) is widely used for glottals because it is in most common fonts and it is more stable than using an ordinary quotation mark in which 'ava'ava is smart quoted into ‘ava’ava. It's main (only?) problem is that custom dictionaries do not accept it as the first letter of a word and they treat words with it inside as two or more words.
U+02BB (Modifier Letter Turned Comma) is the real Unicode glottal and spellers recognize it as a letter rather than a quotation mark. The problem is that U+02BB is in only a couple of little used fonts with XP. With Vista things improved in that v.5.01 (2006) Arial and Times New Roman both contain U+02BB, although it was still absent from most fonts including the new Calibri etc. And if a file in Vista's Arial with ‘Ava‘ava was copied into an XP system it becomes □Ava□ava.
Thus my Cook Islands Maori dictionary of names still has the glottal as U+2018, although when U+02BB is more generally available I will change. Generally all systems handle the macron vowels: āēīōū ĀĒĪŌŪ (although, surprisingly, your Maori_3a lost them).
Back to your various solutions. Maori_2ub and Maori_3a as you noticed do not work - the former is a mess and marked everything with a macron or a U+2018 as misspelt; while the Maori_3a handled within 2018s OK, ignored initial 2018s, and as I note above this solution lost all the macrons.
Solution Maori_1 (for text using U+2018).
This was the Word .dic converted to UTF-8 without BOM.
Macrons: responds to missing macrons (within words and on initial letter) and lists the correct word, eg. Upoa lists Ūpua
Within 2018: responds to missing 2018s and lists correct word, eg. akaoa listed Akao‘a
Initial 2018: marks all words with initial 2018 as incorrect, and for missing initial 2018s such as Ārorangi it correctly lists ‘Ārorangi (which, of course, it continues to mark as incorrect). Not 100% but the suggestion list gives good information to reflect upon and will correct missing initial 2018s.
This is the dic I will be using with Open Office until the day I change to U+02BB.
Solution Maori_1b (for text using U+02BB)
Macrons: as for Maori_1
Within 02BB: as with Maori_1
Initial 02BB: this is where this solution shines. All correct words are shown as correct, and (as for Maori_1) when an initial glottal is missing it lists the correct word.
This soltion will be a great help to those working in Polynesia who have already changed to chr U+02BB - and it is certainly a good reason for me to think more seriously about abandoning U+2018 and changing to U+02BB.
Meitaki ma‘ata
[1]