And there's the hard part. Knowing when to buy and sell. I bought a 7950 for mining (had to get a new PSU as well, so $400) when Bitcoin was $50. I could have bought 8 BTC. Instead, since that time I've mined less than 0.5 BTC, and it's already almost not profitable anymore. I essentially lost 3.5k (as of today's rate) by making the wrong choice.
You are absolutely correct.
The trick I've found is that it is almost always more profitable to buy Bitcoin than to buy mining hardware. The only time it may be profitable to buy mining hardware is when we are on a clear downtrend in price and it is least profitable to mine, and you should only do it if you can get super-cheap mining rates or specialized hardware that gives you an edge over the large datacenter miners.
Now, this might sound counter-intuitive, but uptrends (See GPU mining in 2010, FPGA mining in early 2012) start very slowly once mining gets back into diverse hands that hold the coin instead of immediately sell it. Mining should be unprofitable for reasonably-long periods of time to revert the big-house mining (and immediate selling to cover costs) that happens and put the control back into the hands of smaller-scale operations. I suspect this trend will be necessary for Bitcoin to reach wider adoption and higher acceptance worldwide. Right now the power is concentrated into the hands of too few mining operations, so it is MORE beneficial for the price to fall and individuals operate/mine at a loss and accumulate coin.
So your question of why should you mine at a loss? To help the network and foster the right kind of miner. This topic has come up before in late 2011, and
I addressed it here.