Two calculated tables. One stand alone. One filtered by parent table but returns higher values. Why?

[D_PBI]

Hi,
I've come across something which I'm not sure what's causing it.
I have two calculated tables (a filterable version and a non-filterable version) - see below. They are both created with the exact same formula.

The first calculated table is joined to by '_caseid' from the parent table (the same parent table the two calculated tables are derived from - 'Fund Cases'). Due to this directional relationship, the first table will be filterable.
The second calculated table has no relationships and therefore acts as a standalone table and is not filterable.
The model for this is below. The relationship between the two tables is on '_caseid' to '_caseid'.

The below slicer fields are dragged from the parent 'Fund Cases' table. If filtered, this will reduce the number of caseids records and therefore the directional filtering will reduce (filter) those caseids in the 'Stage Average' calculated table. Note that Stage 1 here has the average of 767.

When I drag the same fields, but from the non-filterable 'Stage Average Unfilterable' table, into the same visual, you'll see the values change. Not all values change. Stages 1, 2, 3 values have changed but Stages 4, 5, 6 haven't. Note Stage 1 value is now 748.

If you view the different datasets, the values from the 'Stage Average Unfiltered' calculated are the same or lower.
Having exported out the raw data and manaully filtered, manually calculated averages, I can confirm the values from the 'Stage Average Unfiltered' table are correct.

I would imagine some implicit filtering is taking place but as I'm a little pushed for time, which preventing me from investigating this in great detail, does anyone have some tips to quickly find the cause of the differing result?
Thanks.