you should try different period to test your model. Such as the period after 911 or the peroid at 2002 when the bubble broke.
In my opion, your model (box) doesn't make any sense.
sorry to say that. There are a lots of portfolio theories. you need to find a way to measure the risk and the return as a whole.