Actual data sample would not do a lot of good - the XML files I work with describe connection and placement of parts so are not too readable (way too complex). I think that Items/Containers example describes basic question well.
To make it slightly more complex lets assume that when an Item is ordered it has color so described by element Item_Color is created and when subsequently that item is placed in a Container it gets a reference to Item_Placement element that has reference pointing to that Container.

<?xml version="1.0" encoding="UTF-8"?>
<Configuration>
  <Description>Items in Containers</Description>
  <Item_Color id="1" color="Blue">
      <Reference refType="Placement Data" id="2"></Reference>
  </Item_Color>
  <Item_Placement id="2">
      <Reference refType="Physical Placement" id="3"></Reference>
  </Item_Placement>
  <Container id="3" color="White"></Container>
  <Item_Color id="4" color="Red">
      <Reference refType="Placement Data" id="6"></Reference>
  </Item_Color>
  <Item_Placement id="6">
      <Reference refType="Physical Placement" id="3"></Reference>
  </Item_Placement>
</Configuration>
[download]

So that means Blue and Red items sit in White container.
The same configuration can be described by a file that has different set of IDs - how do I decide it is stil "Blue and Red in White"?
Is there away to identify that some configuration contains some other as a subset - ie. in some more complex situation we still have "Red and Blue in White" (as part of bigger situation).