I was heavily involved in the beta testing of when DxO introduced keywords - I think it was PL5.
Anyhow, at the same time, I was writing my own image browser, which managed keywords. The main problem that DxO had problems resolving was which metadata tags should be used to be the most compatible with the majority of other DAM apps.
My app has a separate keyword management window, allowing me to organise hierarchies without having to mess around in the very small editing field the PL provides.
Here you can see a small hierarchy that I can apply to an image…
In the image browser, I can select either a whole or part of a hierarchy…
This then gets recorded in the XMP sidecar as…
<dc:subject>
<rdf:Bag>
<rdf:li>Fruit</rdf:li>
<rdf:li>Orange</rdf:li>
<rdf:li>Satsuma</rdf:li>
</rdf:Bag>
</dc:subject>
…
<lr:hierarchicalSubject>
<rdf:Bag>
<rdf:li>Fruit</rdf:li>
<rdf:li>Fruit|Orange</rdf:li>
<rdf:li>Fruit|Orange|Satsuma</rdf:li>
</rdf:Bag>
</lr:hierarchicalSubject>
The rules state that all components of a hierarchy must be written, separately, to the dc;subject tag, along with any other standalone keywords.
Then the lr:hierarchicalSubject tag should contain the full “paths” of all levels in the hierarchy, delimited by |
My app only displays the individual keywords…
… but this is simply a space saver and, as you saw previously, the keyword manager allows you to see them in their context.
PhotoLab sort of mushes this up into a very small workspace, where the only option to building hierarchies is to drag one word up or down the list - hardly enjoyable when the source and target words are at either end of the alphabet.
My manager allows the user to drag words from the right list to a prospective parent in the left list - much easier.
If I now open the image I just keyworded in PL8, this is what I see…
Notice the “hierarchy” list only show the hierarchies used for this particular image.
And, in the DOP file, it is written as one list…
Keywords = {
{
"Fruit",
"Orange",
"Satsuma",
},
{
"Fruit",
},
{
"Fruit",
"Orange",
},
},
Personally, I find this weird but, since I don’t have to unravel it myself, I’ll leave well alone and rely on the XMP sidecar.
Here is what PL8 writes to the XMP sidecar, given the same hierarchical structure…
<dc:subject>
<rdf:Bag>
<rdf:li>Fruit</rdf:li>
<rdf:li>Orange</rdf:li>
<rdf:li>Satsuma</rdf:li>
</rdf:Bag>
</dc:subject>
…
<lr:hierarchicalSubject>
<rdf:Bag>
<rdf:li>Fruit</rdf:li>
<rdf:li>Fruit|Orange</rdf:li>
<rdf:li>Fruit|Orange|Satsuma</rdf:li>
</rdf:Bag>
</lr:hierarchicalSubject>
So, as long as Lightroom writes the same tags to the same keys, you should be OK. What you show in your file in your first post is perfectly reasonable and PL handles it correctly. Apart, that is, from the visual mess 
