Does it make sense to conduct a Hierarchical Principal Compomnent Analysis (PCA)?

So I have twelve variables that theoretically belong to the same dimension and I want do reduce them to one single item. A hierarchical factor analysis gives really low loadings at the second level. However, if I do PCA at both levels I get good results, but does it make sense to do a hierarchical PCA? What are the implications of it? I have read that PCA is only done when you have no underlying theory and just want to see what patterns emerge. However, in my case I do have a strong case why those variables belong together.
I have read in a paper from Welzel and Inglehart 2016, that if one follows compository logic (rather than dimensional logic), it is ok to combine variables that do not load well with each other, if a strong theoretical case can be made. So in that sense, should I just use the factor analysis with lower loadings?
Here is the paper:
question one- is it ok to combine variables if literature supports this: yes. In general, the literature should largely guide your statistical analysis. What I wonder is if there is not a reason that your data does not align with the literature.

And can I ask what you mean by a hierarchical PCA? Do you mean a PCA with adjustments for nested structure?