Query with test:false also excludes regular orders?

[Funnel]

We are submitting a GraphQL query similar to this:

{
  orders(sortKey: CREATED_AT, first: 10, query: "created_at:>='2021-01-01' created_at:<='2021-01-03' test:false") {
    edges {
      cursor
      node {
        name
        id
        test
      }
    }
    pageInfo {
      hasNextPage
    }
  }
}

This results in a response with 6 orders , all with the test attribute set to false. No surprises so far.

If we remove test:false from the query leaving only the date range we get a response with 8 orders, again with the test attribute set to false for all orders.

Seems like the test part of the query is not to be trusted. Did anyone experience something similar?

Thanks,

Martin