One: because people will pay more per terahash while the difficulty is low. That's kind of obvious: your potential return is a lot higher per terahash while the difficulty is low, so therefore AM can charge a lot more. That effect is quickly subsiding, as evidenced by AM's prices.
Two: there is no serious competition yet. Beyond the organic competition anticipated from Avalon, BFL, KncMiner etc getting their shit together, at this point the sales volume of miners is large enough for an established mid-size electronics manufacturer to mosey on over here and embarrass AM price-wise.
Three: electricity cost. Believe it or not, the location of AM (Guangdong province, China) is not the cheapest place for electricity. Don't quote this figure, but I believe it comes out to about $0.06/kWh. As competition ramps up, money getting in on a piece of that 26% APR will launch huge farms in places like Washington state or Siberia where the price is $0.01/kWh - and they will do so by licensing the BitFury chip, for example, not by buying expensive AM hardware. AM will have trouble competing.
All 3 of your points are counteracted by simply being the fastest to release the latest tech miners in bulk.
The real risk is if miners stop buying on the whole, like if bitcoin becomes worth $20 or such.
You think that chip die shrinks are going to maintain the current ratio between BTC spent per miner and coins per month mined? I'm afraid not: the move from GPU to ASIC produced the biggest one-time profit opportunity for mining we will ever see short of quantum computing. And it isn't as if the other ASIC companies aren't working on their own next-gen chips.