Am I fooled by GREP or ...

Question

I have documents with ancient Corel Ventura markers in plain text for index topics. By script I convert the markers to topics, crossreferences etc. The script might look familiair: it's a modified version of index_rebuild by Peter Kahrel. Although the script seems to work quite well, I discovered some odd behaviour.
Two different Ventura markers not seperated by any character are individually recognized if I GREP them manually. After running the script only the first marker has been converted to an Indesign index topic (and can be found in the Index panel).
The Ventura markers look like this: <$I[]Topic;sub topic>
The situation discribed above looks like this: <$I[]Windows 7;64-bits><$I[]Windows 7;32-bits>
Could anyone please explain why the second marker isn't converted?
var myDoc = app.activeDocument;
if (myDoc.indexes.length == 0) myDoc.indexes.add();
var index = myDoc.indexes[0];
app.findGrepPreferences = NothingEnum.nothing;
app.changeGrepPreferences = NothingEnum.nothing;
app.findChangeGrepOptions.includeFootnotes = false;
app.findChangeGrepOptions.includeHiddenLayers = false;
app.findChangeGrepOptions.includeLockedLayersForFind = false;
app.findChangeGrepOptions.includeLockedStoriesForFind = false;
app.findChangeGrepOptions.includeMasterPages = false;
app.findGrepPreferences.findWhat = '<[$][AIS]?\\[\\]([-.\\d\\w\\s]*);*([-.\\d\\w\\s]*)>';
var entries = myDoc.findGrep();
for (i=entries.length-1;i>=0;i--){
    var entry = entries[i];
    var indexType = 0;
    var str = entry.contents;
    if (str.indexOf('<$I') !=-1){
        indexType = 0; // gewone indexingang
    else if (str.indexOf('<$A') !=-1){
        indexType = 1; // See also
    else if (str.indexOf('<$S') !=-1){
        indexType = 2; // See
  switch (indexType){
    case 0:
    var array = str.substring(str.indexOf(']')+1, str.indexOf('>')).split(';');
    var new_top = index.topics.add (array[0]); 
    for (var k = 1; k < array.length; k++)
        new_top = new_top.topics.add (array[k]);
        new_top.pageReferences.add(entry, PageReferenceType.currentPage);
    break;
    case 1:
    var array = str.substring(str.indexOf(']')+1, str.indexOf('>')).split(';');
    var new_top = index.topics.add (array[0]); 
    var new_ref = index.topics.add (array[1]);
    var new_crossRef = new_top.crossReferences.add(new_ref, CrossReferenceType.seeAlso);
    break;
    case 2:
    var array = str.substring(str.indexOf(']')+1, str.indexOf('>')).split(';');
    var new_top = index.topics.add (array[0]); 
    var new_ref = index.topics.add (array[1]);
    var new_crossRef = new_top.crossReferences.add(new_ref, CrossReferenceType.see);
    break;
app.findGrepPreferences.findWhat = '<[$][AIS]?\\[\\]([-.\\d\\w\\s]*);*([-.\\d\\w\\s]*)>';
app.changeGrepPreferences.changeTo = '';
myDoc.changeGrep ();

Answer 1

Your GREP expression is not the problem, nor how the script collects the matches. The problem is a bizarre bug (which I didn't know about). You can see what happens if you comment out the last three lines in the script so that your codes stay in the text:
looks like this: |<$I[]Windows 7;64-bits|><$I[]Windows 7;32-bits>
Here the vertical bars show where the script places the index markers. These markers are always placed at the first insertion point of the referenced object, except when that insertion point is preceded by <>()= and a handful of other symbols. That's why the first index marker is placed correctly right before <, the second one is placed one insertion point too early, before >< instead of after it.
The solution seems to be to place the leading < in a lookbehind:
app.findGrepPreferences.findWhat = '(?=<)[$][AIS]?\\[\\]([-.\\d\\w\\s]*);*([-.\\d\\w\\s]*)>';
Peter