the reality is data points are so limited that we aren't anything close to 'sure' about any ranking system; thats why your method of doing things like "well #6 lost to #4 by 3 points so
it seems right to keep them at #6" is so wrong. I know thats not what is going on here but I wanted to bring it up because it goes to one common flaw with a lot of rating systems.
The reality is we don't know that #4 is #4 or #6 is #6. What we *do know* is that the team in question lost, and that data point has meaning.