1. Competition will always drive miners to dissipate a large fraction of the block reward in electricity. Efficiency gains are only short term.
Long-run Electricity usage depends on the ratio of electricity use to hardware costs. Decreasing this ratio saves electricity (though it does not affect waste overall). You are just replace wasteful electiricty use with wasteful R&D and hardware production. PoW will remain a complete waste regardless of whether that waste occurs via electricity or otherwise.
Nevertheless, you are wrong to suggest that electricity savings are only short-term.
A Minirig SC supposedly yields 50 Mhash/s per $ and 1000 Mhash per joule. The ratio is 200 Joules/s per $ of capital expenditure.
A 5970 yields 1.67 Mhash/s per $ and 2.2 Mhash per joule. The ratio is 1.3 Joules/s per $ of capital expenditure.
As you can see the minirig supposedly has made tremendous progress reducing electricity use, but much less progress in reducing capital costs.
This means that ASIC technology is biased towards energy saving. The long-run competitive equilibrium will feature a quite large reduction in direct electricity use via mining. Perhaps 10-fold.
By contrast, capital expenditure will increase dramatically. This will help to price small miners out of the market. Purchase a mini-rig if they exist, or go home.
I'm too lazy to do all the math at the moment. It is hard to find motivation in a failed technology. Question me further if you want to see exact calculations.
2. Bitcoin is more energy efficient than cash, vaults and armored cars driving around everywhere.
Show math supporting your claim. Currently, bitcoin is orders of magnitude more wasteful than other payment technologies. Here is a reference:
http://www.google.com.sg/url?sa=t&rct=j&q=can%20we%20afford%20proof%20of%20work&source=web&cd=2&cad=rja&ved=0CDcQFjAB&url=http%3A%2F%2Fweis2012.econinfosec.org%2Fpresentation%2FBreuker_presentation_WEIS2012.pdf&ei=ooWwUMzuL8PorAf7tIHYCg&usg=AFQjCNHVP4mza5SJp6E3amVUszxrFR9A2Q. Future circumstances are speculative.