If the target position occurs inside a table, it places the new note immediately before this table. If it finds exactly the same number of superscripts and notes, it moves the text into position. It also circumvents the downright idiotic 'import failure' where you get a blasted superscripted square instead of the right number. (And then you can be pretty sure the right number occurs somewhere before the next note.) Usually, this is not a problem - only rarely I have found it to pick up the wrong number. Any numbers out of sequence are ignored, so in a sequence of text 'note¹, a² + b² = c², note², note³', the first superscripted 1 and 2 are found, then the superscripted 3 at the end. Then it does the same for superscripted numbers. It does so by looking for paragraphs starting with numbers, where each next one should increment by 1. It scans for the exact text '(Endnotes)', which should be somewhere in your document, and counts how many 'notes' it is followed by.
With a script of course - the reverse of Peter Kahrel's FootnoteToEndnote.
