I've wondered about this too...

In this week's Nature:

"Overzealous use of decimal places has wrong kind of impact"
by J. M. D. Coey 23 Sep 2009

"Scientists teach students to evaluate critically the significance of their measurements, and to eschew meaningless decimal places thrown up when pocket calculators work out a quotient of two integers. So what are we to make of the recently released impact factors, including Nature's much advertised rating of '31.434' (see also http://www.nature.com/nature/about)? Has Thomson Reuters discovered a protocol that allows it to measure the impact of a journal with an accuracy of 32 p.p.m.?

Quoting this figure conveys the wrong impression — that innumerate marketing is trumping common sense at the heart of science's leading journal."

So what does a thousandth of an impact factor rating look like anyway? Also, how much does Thomson Reuters charge per point? ^_^