My results agree with yours, in terms of who wins. Python 2.6 (2.4 is similar, and 3.0 is as well, for the techniques that were still valid) on OSX 10.5, Intel Core Duo at 2 GHz:

  [scotts@silver]$ /opt/local/bin/python2.6 f.py
  f1 0.107
  f2 0.19
  f3 0.106
  f4 0.097
  f5 0.139
  f6 0.073
  f7 0.025
  f8 0.103

edit: I added a ninth function,

  def f9(list):
    return ''.join(map(chr, list))

To test if the difference between f8 and f6 was the list comprehension or the joinfields method. A result of 0.072 tells me it's the list comprehension. The only difference between f8 and f9 is list comprehension versus map. That leads me to conclude that map is indeed a faster way to construct a list than a list comprehension. (Although this may change for a user-defined function. I could test this, but I've spent enough time already.)