I think your problem is that m//g doesn't grab overlapping matches without some help from you. In scalar context, it picks up where it left off.

After a successful match, determine by inspection where the matching term is in the string, and then set pos just after that. Perhaps something like this (untested):

while ($text =~ /$expr/g)
{
  my ($prefix,$term,$suffix) = ($1, $2, $3);
  print "$prefix<b>$term</b>$suffix\n";
  # update pos to just after $term
  pos() = pos() - length($suffix);
}
[download]