The best way is by an example:
//echo -a $regex(ab cd ef,/([a-z]+) \1 \1/)
you can see why this fails: \1 is filled with the captured string from ([a-z]+), which is "ab". So this regex is trying to match whatever was first captured 3 times (ie "ab ab ab") which of course fails.
Now consider this:
//echo -a $regex(ab cd ef,/([a-z]+) (?1) (?1)/g)
this succeeds because (?1) is actually a shortcut for the first () subpattern in the regex. So the above pattern is equivalent to
/([a-z]+) ([a-z]+) ([a-z]+)/
or more accurately to
/([a-z]+) (?:[a-z]+) (?:[a-z]+)/
since (?N) themselves aren't capturing subpatterns.
In markup1, the second red part is a shortcut for the (much lengthier) first part. That second part is what's repeated, thus saving characters.
