Just for fun, with all this talk about flashlights, minimum lumens, etc., I decided to flip this question on its head.  (See how I spend a vacation day?:))
My question is what's the minimum number of lumens the human eye can actually see?  That's the flashlight I want!  

We know the human eye can reliably respond to an input of 90 photons arriving irregularly or a single photon arriving regularly, which we'll say reasonably corresponds to our definition of a flashlight.  (D.A. Baylor, T.D. Lamb, K.W. Yau, "Response of retinal rods to single photons."  Journal of Physiology, Lond. 288, 613-634.  1979.)
So, how to translate that to lumens?
Small numbers of photons will be detected scotopically by the rods in your eyes (as opposed to the cones), and each rod has a diameter of approximately 0.002mm, giving each an area of 0.0063mm^2. 
Now, let's consider a 
1 photon flashlight, sure to be coming to a store near you real soon now.  (Can we get a Doug Ritter model???)
We'll suppose the photon has a frequency of 540nm, which would make it "greenish" on a larger scale.
Then, the rod would have an incidence of 2.39x10^13 photons per second per lumen.  Working backwards, we have that a single rod can detect 4.18x10^-14 lumens, assuming the flashlight were outputing one photon/second.
And there ya go.  So keep yer fancy 200 lumen lights!  I'll be more than satisfied with my 
0.00000000000004177 lumen model!  
 
  
 
  (It probably runs for 100,000 years on a single battery.  And if you're willing to use a radioactive battery, it can run for millions of years on a single "charge."  Try that with your Surefire!)