Whatever else this thread may or may not prove, what it should do is cause everyone to take a closer look at the analytics data that is available and see if there is any wisdom therein to be gleaned.
I spent a while today exporting the type of data being discussed here and mixed it up a bit on a spreadsheet. I used 12 months of data from immediately before my site's traffic started to decline.
One useful element that you have to calculate (as it is not isolated in analytics) is the number of "internal arrivals" (that being Unique Page Views minus Entrances). This tells you whether the page is more of a landing-pad for searchers or a link-in-the-chain for people progressing across your site. The actual character of the page should match your expectations for the page.
I then went for the simple stuff and looked at the pages with the lowest page view numbers. This isolated a group of pages that obviously were not working well, they were not landing-pages, they were not links-in-the-chain and they were not destinations at the end of a chain. These will be deleted, but the decision to do this is made based on page views rather than exit rates.
Staying with the simple factors I next looked for lowest time spent on page. I was gratified to see that these were mostly category "link-in-the-chain" pages with a high percentage of internal arrivals and a low exit rate. (I concede that a high exit rate here combined with high internal arrivals would have been a red flag warning.)
Finally I looked at my highest exit rates. These turned out to be mostly articles and directory pages. Typically these have a 70-80% exit rate, but the time on page is around 2-3 minutes. They have an internal arrivals rate of about 16%. So these pages are landing pads that serve the searchers' purpose well.
So I've used my data in a less complex way than claaaarkey, but that may be because my site is less complex than his.
If there's a rule-of-thumb to be coined here I think it would be this: As long as your higher exit rate pages also have a high time on page and low internal arrival rate, you're probably doing things right.