We currently support cross-temporal reconciliation via the iterative approach, but this breaks coherence when done for few iterations (and is unclear from the current documentation)
We need to support cross-temporal reconciliation via the full projection approach too.