In fact the whole setup is much more complicated than this test case.

We are talking about a query that uses about 5 tables.
Here are the number of relevant row per "test"
table 1: 1 row
table 2: 8 rows
table 3: 125 rows
table 4: 624 rows
table 5: 2197 rows


table 1 => table 2 => table 3 => table 4 => table 5

the data is not really duplicated as it is calculated, it is a mix of avg, sum, max, min, .. then consolidated into a "flat" table for reporting.