Results mismatch in two similar sDA results

Hi Team,
I ran a sDA assessment for two cases - with and without a courtyard, with 5 ab.
However, in the two cases the top room performs differently, although it is not effected by the courtyard.
It would be good to know how this difference can be minimized ?

PS. I ran the same model in IES and there it gives a difference of only 0.2% in the top room.