(dial.) Luncheon, lunch.
An excursion into etymology...
A nuncheon is a light meal in the middle of the day, that is it's the same as lunch or luncheon. So obviously the words must be related, yes? Well, yes and no, and not at all in the obvious way. Webster 1913 has an entry for the spelling nunchion, but Webster 1913 is a ratbag, and nuncheon is by far the more familiar spelling. Further, it gives a nice comparison with luncheon. (Actually, other spellings for both existed in the past: lunchion, lunshin, etc.) And since Webster's etymologies can be rather hard to read, I'm going to lay it out more
clearly fully here and adduce some linguistic points.
In origin there are two totally different words. By chance they came to mean very similar things, and had similar forms, so one form influenced another and made it even closer.
Lunch first. In origin this seems to be a variant of lump. We have several other such pairs: bump and bunch, hump and hunch. There is no known phonetic reason for these; it just seems Old English or Middle English (depending on how far back they go) made a habit of using alternating pairs like this. So a lunch began as a lump of something, like a lump of bread, to keep you going out in the fields between meals.
Next came luncheon, which was probably* an extended form of the word lunch. There are other pairs like this too, such as punch and puncheon. It's probably a Latin or Romance augmentative suffix: much the same as made balloon from ball. Now we would naturally think that lunch is short for luncheon, but it seems to be the other way round: luncheon is long for lunch.
Nuncheon is a compound, of noon and a word that no longer exists in modern English. Webster lists it under skink but it's completely obsolete. Another form of it was shench. So an earlier form of nuncheon was something like noonshench.
A shench or skink was a drink. The noonshench was the drink you had with your noonmeat, at noon. Another word for noonmeat was of course lunch or luncheon. At some point both noonshench and luncheon came to refer to the meal itself, rather than something you had at it.
Now, sometime in the Middle English period (you could date it better if you had detailed citations, but I've just got a couple of smaller etymological dictionaries), the word noonshench fell into the gravitational clutches of the word luncheon.
The change of vowel from oo to u is not exactly regular but I'd imagine other examples could be found. The change from nsh to nch is a minor phonetic detail, and quite natural. The extraordinary thing is the loss of that final ch. There is absolutely no phonetic justification for this at all. This never happened in the history of English: bench didn't become *ben, and lunch didn't become *lun.
But the two words were now so close in meaning that speakers must have perceived them as almost variants of one another, and the mighty force of analogy overruled the laws of phonetic regularity, and dropped the final ch so the two words would match. It's a common process in language; but each time it happens it's different, usually applying only to a single word like this.
We can take the two parts of noonshench back further. Noon is from the Latin ordinal nona, 'ninth', that is the ninth hour from the beginning of the day. The calculation of the hours of the day changed since it got its name, but noon and the nones kept their old names.
Now a shench was a drink, an obsolete word, but apparently surviving in the dialectal skink, a drink, or a cup or goblet, or to pour or serve drink. This originates from the means of pouring: ultimately related to our word shank. The core meaning is the bone, thence a hollow bone used as a pipe, then the sense of serving drink through a pipe.
My dictionary tells me there's also a dialectal word skinker, meaning a tapster, one who pours out liquor. I haven't heard of this. But the root is more prominent in German, where the related word is Schenk, meaning a cupbearer or publican. From this comes the everyday word schenken, which in modern German is the ordinary word for 'to give (someone a present)', quite devoid of any specific connotation of drink (or pipes or bones).
Why sh- and sk-? It was originally sk-
in Proto-Germanic, and so entered Old English, Old High German, Old Norse, and its other descendants. In the Old English period the sk- underwent palatalization, becoming sh-. It did the same in German, though later than in English: the spelling Sch- preserves the C, indicating that it had become fixed before the change was complete. But our word skink (the drink, not the lizard) seems to have come from some form of Norse (Scandinavian), which didn't have this sound shift.
And why -k and -ch in skink/shank/shench? Another palatalization in Old English. There are other such pairings: drink and drench, bake and batch, stink and stench. It was a fairly regular phonetic alternation in Old English. So we have the alternation in the native words shank/shench; and in addition the borrowed skink from Scandinavian, lacking the sound shift.
I was thinking of consulting the OED tomorrow for the history of forms by date, but WaldemarExkul kindly supplied me with the entry. It is very complicated! What I've presented here is way oversimplified, though essentially correct.
* I've had to add this 'probably' here. The evidence isn't clear-cut, and someone could come up with better ideas or better evidence.
I've now heard two alleged Spanish origins, lonja and las once. The first is possible on phonetic grounds, the second impossible; but you can always find meaningless coincidences between any two languages. Linguists work by Neogrammarian principles and demand explanations of every single sound change, and lexicographers demand written evidence.