Your analysis is flawed.  The most "efficient" way to pay has nothing at all to do with the power consumption, but everything to do with the costs of using the system to the consumer doing the paying.  In this context, "efficient" means "cost effective".  There is no doubt at all that Bitcoin is orders of magnitude lower in system costs for the end user than Visa.  I don't even need to demostrate that one.  Bitcoin is also significantly lower than the same metric applied to paper currencies traded in hand, but that comparision needs a bit more explaination.

While bitcoin's total energy consumption seems high, it's also relatively easy to get a pretty reliable estimate of what the system costs to run.  The total costs of running a paper currency system are more difficult to estimate, in part, because so many of those costs are obscured from the users.  Most are paid for by inflation and/or taxes, so users don't bear those burdens anyway; some such costs include (but are not limited to) The costs of manufacturing the paper and printing the currency itself.  The costs of the beauracy required to administer the currency & regulate the banks.  The cost of enforcement of currency related crimes (i.e. counterfitting) that would otherwise be irrelevent under a cryptocurrency, plus all the associated court costs, prison costs, attorneys, and social costs of prosecution; and so on.  And to that, the costs of the people involved in the handling of said currencies (outside of the beauracy); for example, cashiers and managers who must be trusted with such funds and the labor time required to properly manage such funds for every retail store, compounded to every retail businees in the country with a cash machine.  Then there are the armored truck companies, banks and such institutuions that exist for the sole purpose of accumulating, protecting and accounting for other people's money.

And then account for the costs of heating and air conditioning the buildings all those institutions must maintain in order to comfortablely house and employ all those bankers.

And I haven't even touched on the transaction costs that consumers commonly bear directly; such as the gas and time required to travel to a bank on a regular basis to conduct their personal financial business, nor the costs and risks involved in distance commerce using paper based currencies, such as mail order scams or online scams.

There is no such thing as a free lunch, so no matter how the costs of any paper currency are officially paid for, the burden must eventually fall upon the consumers.  The finance industry in the US is now one of the largest industries recorded in the GDP, and by far the largest that does not actually produce a tangible product or offer a free market service of it's own.  If you take the slice of the GDP that is attributed to the finance industry, and that is how much the system costs overall.  

http://www.huffingtonpost.com/2011/12/15/financial-sector-economy_n_1151058.html

Divide whatever number you decide upon there by the current number of American households, and you come up with the appproximate cost of the system for each household per year.  Not exact, mind you, because some people end up bearing more of the burden than others due to the high externalities of the system (by design), but it's a pretty good number since most of the burden falls upon the middle class anyway.  Once you've done that calculation to your own satisfaction, do an estimate concerning how many individual transactions that would actually be per year per household, right down to the candy bar at the corner store for little Jimmey.  That's going to be a pretty high number; but even assuming that all such transactions remained individual, near-time fee paying transactions on the blockchain (an unlikely assumption, but the worst case scenario for our Bitcoin) and times the current minimum fee.  (what is it now, roughly half a cent, right?)  See for yourself which would be cheaper for you individually, and for society collectively.