OK, so they have messed things up a little bit. The collation is for code page 850, and in code page 850 0xD1 is indeed "Đ". But if they turn off translation when sending data to the application, the application may still display Ñ. However, if they sort the data, what looks like MUÑOZ would sort between MUCHO and MUERTO.
From this experiment, we can sense a workaround for this would be
SELECT convert(nvarchar(40), convert(varchar(40), binary_value))
FROM OPENQUERY(SERVER, 'SELECT convert(varbinary(40), col) AS binary_value FROM tbl')