[CS4] !!text Hebrew with problem unexpected

Question

This is driving me !sreknob
I have a couple of Hebrew phrases in an otherwise English text, so I assigned the World Ready Composer to the styles in question using Peter Kahrel's "Arabic" set of scripts, and created a 'Hebrew' character style. No problem: the Hebrew text comes out perfect. What's Not So Nice is that some other text also follows an RTL direction!
In the above screenshot, you can see the volume number and page range "2:663-676" appear correctly in the Story Editor below. The Story Editor does not know anything about Hebrew, so it reverses the Hebrew text. That's not a problem, because the World Ready Composer fixes that in the formatted text (above).
But what is going on with the numbers!? No amount of copying & pasting seems to fix this. The paragraph direction is set to Left To Right; the Hebrew text has a character style that only applies another font.
[Edit] Oh, and of course the square bracket at the right of the numbers isn't corerct either. ]It seems more than a bit [random to me.]

Answer 1

A-ha! I was right, it's an automatic system going wild. Thanks for confirming it!
> Sometimes it's easiest just to retype the whole thing....
I must have used at least half a dozen different methods in the past. Fortunately, each one more advanced than the previous! Let me think, I have:
* printed from Word to PDF and painstakingly imported/placed/scaled each separate phrase. Pro: anyone can do this. Con: Loads of extra work. Loads of fun with line-spanning text, and the occasional last minute correction. Cannot adjust font.
* copied from PDF into Illustrator, saved as .ai, the rest as above. Pro: you can tinker with the text (a bit). Con: even more work.
* reversed the characters in ID. Pro: fast, efficient, can use fonts. Con: it qorked nicely in CS3 and older, but CS4 tries to override accent placing and ligatures, which is troublesome for Hebrew. Also, Arabic needs to have Each Bloody Separate Character checked for correct initial/medial/final form. Another con: line-wrapping needs to be done manually. Only just feasible with short phrases; longer stories -- think of a full paragraph -- needs to be broken manually into lines, because otherwise right-to-left text also appears bottom-to-top.
* cut phrases out of running text, pasted into separate text box, so I could finally use proper RTL character and paragraph formatting. Text boxes were then re-inserted into the text on the original position. Pro: finally! Automatic ligatures, wonky text cursor, real fonts -- the Real Thing. Con: still no good solution for line-wrapping text. Also, the text boxes themselves caused spacing problems.
With all said and done, the following worked out nicely at the end:
1. Add WRC to the correct styles. (I am *neurotic* about getting my style inheritance right. Since in this article both Hebrew and Arabic text appeared in different situations, I could set the WRC to only one style, and the rest also got it from there.)
2.  Select all text and blindly set its character direction to Left-to-Right.
3. Search for continuous spans of Hebrew and Arabic (which may include spaces and the occasional punctuation) and apply the right character set for the script to this. At this point I could either include or exclude spaces, periods at end, etc. by manually verifying if the GREP search matched the correct phrase.
At a couple of positions I encountered the "Operator Error", where a single space was *not* included inside a long RTL phrase -- but only because in a case like this I do not trust my senses nor InDesign, and verify Each Seperate Phrase against the original copy ...