tedster - 7:17 pm on Jan 7, 2012 (gmt 0)
Bounce rate (as determined by standard analytics packages) is a notoriously challenging metric. I've seen a high performing page with a bounce rate of 85% - and this was a page with 750,000 search visits per week from 4 very competitive query terms!
When that page's traffic was confined only to Google search traffic, the bounce rate was still 82%.
So it seems clear to me that, as Matt Cutts and others have said, bounce rate is too noisy a metric to lean on very heavily.
I've seen a lot of analytics over the years, and I'd say both those numbers are very low, but it's hard to make any decision based on such a number. I think the search engines tend to look only at bounces that go back to the SERP within a handful of seconds and then make another choice from the same result set. That is sometimes called a "fast click". And even then, we still have a relatively noisy signal.
Average time-on-page is just a bit less noisy, IMO - and for that reason a higher time-on-page has more appeal to me. That still assumes that the extra time is because the layout is friendlier to the eye and gaining more engagement. It's always possible that a layout is more difficult to cope with, and that might also increase time-on-page.