models suck if the assumptions are not valid!

in your case, these are the issues I see after a quick glance:

-no model for fluctuation of BTC prices
-assume exponential growth of total network hashing power for the next 40 increases
-difficulty increase is not based on a model for expected hashing power but rather on a static +50% increase
-no link between price of BTC and difficulty
-where is the time period?  39 difficulty increases is more than a year's worth of mining...