I agree with John and Neil that time metrics, like any metric, are
mainly important when given meaning and they have little meaning
until you correlate them to success, failure, or something actionable.

As an advertiser, if my boss went nuts and required me to use a
site's Time per Visit or Time per Page to determine my spend without
a correlation between those and conversion, my gut instict would be
be to shy away from both long and short times. Short because it
might indicate high bounces or just because viewers might not have
time to perceive my ads.

I'd stay away from very long times for an even larger variety of
reasons. A high per page time means that, unless my ad floats, the
viewer has spent most of the time /not/ perceiving my ad. A high on
site time might indicate the site is /too/ engaging or sticky, say a
game, so the viewer is unlikely to notice or click ads. Or that the
site is the kind you leave open in a tab in the background all day,
like gmail.

Either way, Neilsen(or someone) will need to show both that time
correlates to "value" and that it does so better than the pageview or
the visit.

--
Brian Monaghan