I am currently working with the Paradise Papers sandbox offered on your website.
After analyzing the data for a while, I'm pretty sure you have a mistake in the scripts setting up the sandbox:
If I'm not totally wrong, you accidentally import the 'Intermediary' nodes as 'Other' nodes and vice versa.
According to your sandbox, the Paradise Papers contain roughly 2000 nodes with label 'Intermediary' and 200 nodes with label 'Other'.
If I compare these numbers to the Neo4j Desktop provided on ICIJ's website and restrict it to the data with source 'Paradise Papers - Appleby' (which should be the dataset you provide, maybe except for some minor updates), it's exactly the other way round: roughly 2000 nodes with label 'Other' and 200 nodes with label 'Intermediary'.
Looking at the edges connected to each of the two labels, swapping the labels would make way more sense:
According to the ICIJ, nodes of label 'Intermediary' represent a middleman or law-firm founding an offshore company on a client's request whereas nodes of label 'Other' group nodes belonging to the same company / holding.
In the sandbox, the vast majority of the edges connected to a node labelled 'Other' is of type 'Intermediary_Of' or 'Officer_Of' which seems strange for nodes whose purpose is just grouping.
Looking at the nodes of label 'Intermediary', most edges are of type 'Connected_To' and not a single one of type 'Intermediary_Of' which seems again strange for a node labelled as an Intermediary.
Simply swapping the labels however would lead to a different and way more realistic semantic:
'Intermediary' nodes would then be connected using 'Intermediary_Of' / 'Officer_Of' edges and 'Other' nodes would use 'Connected_To' edges to group nodes.
I hope someone can confirm my observation (and interpretation) and if so correct the data import for the Paradise Papers Sandbox.