I decided to look at this problem from a different angle this morning and only focus on words that are scribal corrections. I think this gets at the core of the problem better. The approximate counts where corrections were made (including endings):
ο/ω correction: 22 times
α/ο correction: 15 times
α/ω correction: 3 times
As a control, I looked at the common ει and ι swapping:
ει/ι correction: about 100 times (about 1/4 of the way through I got tired of counting and extrapolated the rest)
Again, these α/ο and ο/ω do not appear to be near the order of magnitude as the ει/ι correction which can occur on about any word, by about any scribe. This still does not show the whole picture, however, because different letters occur with different frequency. Here are the letter frequency counts from my raw data:
So again thinking about order of magnitude, with ω being a lot more rare, the ο/ω correction becomes a lot more significant, closer to the magnitude of the ει/ι correction. The α/ο correction becomes less significant as those letters are a lot more common. And the α/ω correction is just dribble. But because α/ο did have a good number of corrections, I am still wondering about my continuum theory of α < ο < ω with ο being closer to ω than to α. Yet for my purposes, the lines have to be drawn somewhere. There are also other repeated types of swapping such as α/ε and υ/ι swapping, but not at what I would consider to be significant levels. I think these percentages are clearing it up for me. Thus, I am inclined to stick with the α != ο and ο = ω rules as being reasonable for now (and it makes it easier for me), but would like to see the other data on the matter if it has been published somewhere.