While I won't speculate either way on whether or not these are a good buy, pointing to the likely (much) higher future difficulty is only half of the equation.
A lot of people have already made an absolute killing on bitcoin by mining when others told them they were stupid because it wasn't profitable. A few years ago, when bitcoins cost a couple pennies each, people would run the same numbers and call those that chose to mine incapable of doing basic math.
Don't forget that a lot of people are mining today based on their own assumptions about the future price of cryptocurrency. Sure, the math doesn't look great based on ~$100 per BTC, but what is you assume bitcoin goes mainstream and hits $1000/BTC in a few years? There will only ever be 21 million coins, and it bitcoin becomes say, the internet currency of choice, then $1000 per BTC is quite a low estimate.
Don't forget that the USB miners are priced in BTC. If bitcoin becomes the internet currency of choice and goes to $1,000, then you would have been better keeping the 2BTC that the USB miner costed, instead of spending them to get back 1.5BTC after 2 years. Pretty basic math.
I would say more: even if the miner is priced in fiat, I always calculate how many BTC I can buy in that moment with that amount of fiat. I decide then if I believe that the miner will generate more than that: if it's the case, I will buy the miner. If it's not the case, I will just buy the BTC.
It seems the only valid logic to me, but I read posts after posts repeating what you just wrote. I really don't get it. Buying a $200 miner when the exchange rate is 1BTC = $100 to get back 0.5BTC is a net loss in my book, regardless of the future exchange rate.