I think the advice given to me in a previous answer is appropriate here.
There comes a time when umpteen bits of keyhole surgery just aren't getting anywhere - then it's time to lift the bonnet, winch the engine out and take the whole thing apart.
You can try searching for Imlach or Cumming as keywords in combination with Taylor (and Jude) but in that case you are still searching for names, not records. That will limit you to records where
- the records have been indexed
- the names you want have been indexed
- the names you want are spelled the way you're expecting
You don't know the place your person is from, so you can't use the techniques you would use for a place study—or can you? In this instance, you can study the military unit that John Jude Taylor served in to see if you can find any clues.
How was the 65th Foot Regiment formed? How were its members recruited? Did they come from a particular geographical area, or were members assigned to it from elsewhere? Can you find unit histories, or letters and diaries of people who served in it, people who might have served with John Jude Taylor?
Have you thoroughly researched Ellen Morrison and her family? At the very least, you could see if the names Imlach or Cumming came from her family and not his family (or from any of his military comrades).
Do John Jude Taylor and Ellen have any siblings?
Do John Jude Taylor and Ellen have any siblings? Have you thoroughly researched all of the couple's children?
How much do you know about John Jude Taylor's military service? Did he have any special duties that might be a clue to his prior life?
How many people in their community in New Zealand also soldiers in the 65th Foot?
Since you are researching in New Zealand, I would start by searching the newspapers at Papers Past and also at Trove for articles about the the 65th Foot Regiment and the colonization efforts in general. Search widely in time to catch retrospective articles that may have been published long after the events took place. If you have access to the newspapers at the British Newspaper Archive (either on its site or via findmypast) search there as well.
Many searches fail because we search too narrowly in time. I have found clues about research subjects as much as 150 years after the events described in the article.
Don't just look for individual people. Think about the Big Picture, the whole process of how the 65th Foot was gathered and sent to New Zealand. Read for context, and you'll be more alert to clues in records that you might miss otherwise.
If you can find rosters with the names of other members of the 65th Foot, and researchers who are studying those soldiers, look at their research. What sources did they cite to back up their research? What records were they able to find, and how did they use them as part of their proof or research?
Go outside genealogy sites and look at the work of local historians, of people who do population studies, of military historians.
A final note: when you are looking for combinations of surnames in keyword searches, do keep a log of what searches you tried and what the results were. Record negative results, and the results of your evaluation of the databases (e.g. "this database had 2400 entries where the names were not indexed"). Write down your thoughts so you can go back later and review your prior work after you have learned new facts from your readings for context. Reviewing prior work should always be part of our research process, but it's especially necessary when working complex cases.
Use whatever record-keeping works for you. I have Scrivener and Scapple, Evidentia, and Clooz in my research toolkit for cases like this. Ordinary lineage-linked software is designed to keep track of information where you've already reached a conclusion, so it isn't always helpful when you are conducting wide searches and haven't reached a conclusion yet. The newest version of Clooz, Clooz 4 supports "businesses, buildings, ships and artifacts" as research subjects, so you could use it to track the 65th Foot as a research subject.
Resources: