The Graveyard of Dead Gods

Christopher Hitchens (and other lesser known debaters) are fond of pitching an argument for metaphysical naturalism which was first popularised in an essay by H.L. Mencken in which the author enumerates the various and sundry gods once beloved of man, and leaves the remainder of the problem as an excercise for the reader.

The problem with this approach, of course, is that it isn't really an argument which leads validly to its preuspposed conclusion.  As it happens, this no easy trick to pull off, and one greatly risks tripping across the genetic fallacy in the process.  Nonetheless, I think that a valid argument may be lurking in here somewhere.  Here is one attempt:

1. If humankind generally has a strong propensity for making up immaterial minds with magical powers, then probably all concepts fitting that description are merely made up.
   
   
2. Humankind does in fact have just such a propensity, and in great abundance.  
   
   
3. Therefore, all allegations of  immaterial minds with magical powers are made up.

Mencken and his latter-day exponents primarily work to buttress premise #2, by demonstrating the overabundance of gods which are today universally acknowledged to be naught but fictions.  

However, the problematic premise here is the first one.  Is it right to make the inductive leap from a general propensity to play pretend with immaterial invisible imaginary minds to the conclusion that all such things are made up?

Imagine that your five-year-old child has an imaginary friend, with whom she speaks and frolics, and who occasionally makes bizarre demands.  You would almost certainly dismiss the notion that her friend is really there, even if you get the sense that the friend is really quite compelling to your child.  Why would you do that, if you doubt the inference made in premise #1?  Is this not precisely the reason that we adults assume that the invisible playmates are not, in fact, direct apprehension of otherworldly spirits?