The only difference between positive and negative lookbehind is
whether a match is treated as success or failure.  They're both
supposed to exhaust all possibilities trying to find the match but, as
you've discovered, they don't.  Below is a test case I wrote up.  All
four regexes should match "foo1", "foo2" and "foo3", but only the
fourth one does, and it's a hack.

import java.util.regex.*;

public class Test
{
 public static void main(String[] args)
 {
   String str =
     "%foo1\n%bar foo2\n%bar  foo3\n%blahblah foo4\nfoo5";

   String[] rgxs = { "(?<=%.{0,5})foo\\d",
                     "(?<=%.{0,5}?)foo\\d",
                     "(?<=%.{0,5}\\b)foo\\d",
                     "foo\\d(?<=%.{0,5}foo\\d)" };

   for (int i=0; i<rgxs.length; i++)
   {
     Pattern p = Pattern.compile(rgxs[i]);
     Matcher m = p.matcher(str);
     System.out.println();
     System.out.println(p.pattern());
     while (m.find())
     {
       System.out.println(m.group());
     }
   }
 }
}

Btw, this one should also do the job if you just want to omit a single
char:

^[^%]*pattern

Kind regards

   robert