G'day Draigcoch, and welcome to the forum.
The meaning of the line in its current format is clear to me (although I accept I may be alone in this - see below): the troops who are not Ax (S) can only be of one type - Ax (O) or Ps (S).
From my understanding of English grammar and syntax, the second "all" is unnecessary. I read it as
...remainder all {Irr Ax (O) @ 3ap} or {Irr Ps (S) @ 3ap}
That is, the word "all" by itself should be sufficient to explain that the entire selection comes from one category or the other.
And now to the point I was referring to above: Rule books like this need to be clear to avoid disputes such as this one raised by Draigcoch. The fact that the point was clear to me is irrelevant if most readers are in the same position as Draigcoch. Unfortunately the list books are full of inconsistent approaches to issues, leading to any number of problems. I raised this on the DBMM email list a couple of months ago, pointing out some examples and proposing some solutions. For example, date limits are expressed using two mutually contradictory systems: options may be available
before a year or
until a year, and they may be available
from a year or
after a year. If you look at the Early Burgundian list in Book 4, you'll see that Swiss Pk (S) are available after 1464, and that Charles the Bold can be downgraded to Inert from 1465; a moment's thought will show that the two lines have identical date restrictions. Add in issues such as whether there are any differences between "replace" and "upgrade", inconsistent formatting and silly typos and it becomes apparent (to me at least) that any project to review the lists would benefit greatly from appointing an experienced editor to develop a style guide and oversee its application to the books.
Returning to the original point, if this discussion doesn't go anywhere, it might benefit from being raised on the email list. It's the sort of example which might help reinforce the point I made on the list. And it might help improve the signal to noise ratio (the poor state of which was the reason I left the list recently).