The reason is this: there are hot spots in the mtDNA which tend to mutate
much more rapidly (an order of magnitude or two) than the rest of the
sequence. So suppose we have two groups for which we are trying to date
a common ancestor. The old technique was just to count mtDNA differences
and divide by the clock rate to get the divergence age. This is now
looking overly simplistic--the hot spots may have mutated several times
(almost a gauranteed reality) meaning that there are two clock rates to
worry about: the hot spot rate (which is good for ages up-until-comparable-to
the expected time of second mutation), and the rate for the rest of the
sequence. For long dates (longer than the expected second mutation time in
the hot spots), the hot spots must be ignored (since nobody knows how to
tell what mutations have gone on in a place*). But in the old regime,
these hot spots were counted, and due to the nature of DNA, they'd probably
be mutated differently in the two groups (just by chance), and so would go
into the (erroneous) mutation count. I'm not sure how much difference
adjusting for hot spots will make, but there are lots of people who
would be happier with longer ages than what molecular clock dating has
given--archaelogists and anthropologists have been surprised, it has seemed
to me, that human divergences were as recent as they were (i.e. some 30-40kya
for human immigration to Europe), and a back-adjustment of the clock would
coincide better with other lines of evidence placing the events further
back. There have been other similar surprises among paleontologists.
Perhaps a reconsideration of molecular clocks (the mtDNA as well as others)
to take hot spots into account will yield a tighter convergence between
the results they give.
* There are some other complications--it is possible to get multiple sets
of mtDNA in one's mitochondria (indeed, this is how the whole thing seems
to have gotten started), and when this happens, you can begin to trace out
a route of mutations in the hot spots. This could (in some lucky cases)
serve as sort of a 'second hand' on the molecular clock allowing finer
dates to be established.
-Greg