May I suggest that for debugging only, separate each rule to a different audience, and verify each one independently. Then create another audience to include members of each of the individual audiences (OR rule as you mentioned). Then check the total.
If all is OK, you can then build a single audience using OR between rules, plus I presume AND within each rule. Then remove the test audiences. The use of 'include children' while selecting items is particularly important for organisation levels/locations.

cheers, Andrew