I received this from Jon Mided at digitaltruth.com:

To settle the original question: "What does 'Improved Midtone Separation
mean'?" -- it seems it does indeed mean _lower_ contrast.  Even though
I agree with J-D that greater tone separation should mean greater contrast.

There are some of us who will only believe a sensitometeric test stand
and densitometer results and those who rely on subjective impressions.
Myself, if I can't measure it then it isn't there -- an engineering
attitude, but do you want to drive over a bridge that is only supported
by belief?

From Jon's description it would appear a claim is made for an H&D
curve that tends to be 'chair shaped': a lowering of mid-tone contrast
while high and low tones are left at normal contrast.

It would seem that two contact prints through a step tablet and a
few minutes at the densitometer would show what is happening.

"Oh! let us never, never doubt what nobody is sure about!"
                                           Hilaire Belloc

"Only issues that can't be settled are worth talking about."
                                                         Me

  In general I have to agree with this. The relation
between the tones of the original scene and the tone
reproduction on the print are controlled by the response of
both emulsions. The response is what is shown in the
characteristic curve, sometimes called the H&D curve. These
curves vary from paper to paper and from film to film. Note
that overall contrast is shown by the slope of the curve.
High contrast materials range from their minimum to maximum
densities with small changes in exposing light, low contrast
materials require a large change of exposure for the same
density variation. However, the _relative_ values depend on
the distribution of sensitivity of the halide particles,
which is what the _shape_ of the curve reveals. A
demonstration can be found by comparing two Kodak films:
400T-Max and ISO 320 Tri-X. T-Max is a relatively "straight
line" film with a short toe (definition of toe is below).
The ISO-320 version of Tri-X has a very long toe, in fact
the curve is upward deflected all along its usable length.
The difference in tone rendition will evident by thinking
about the effect of this in relation to the original scene
brightnesses. If you match the two curves for equal contrast
index and for the same shadow and highlight densities, you
will find that the Tri-X has less density for the mid-grays,
so it will reproduce them darker than T-Max. This is
sometimes desirable. One can find similar differences in
paper curves except that a long toe on paper will result in
brighter mid tones.
 In general, developers have little effect on the shape of
the curve. However, additives like Potassium bromide does
affect them. Bromide tends to suppress the very lowest
exposures, so it has the effect of reducing the usable toe
area. Another way of stating this is that it increases the
relative contrast of the toe of either film or paper. The
effect on film is to suppress fog and, if much is used, to
lower speed. On paper the effect is also to reduce fog and
to brighten the highlights.
  Specific developing agents do not have much effect but
overall contrast of the paper can be varied a little (for
some papers). The variation possible is not nearly so great
as for film because film for pictorial use (as opposed to
graphic arts use) is developed far short of its maximum
possible contrast. Prints, OTOH, are usually developed to
their maximum contrast, or at least, to reach their maximum
densities. Some so called low contrast developers are merely
slow and do not reach this density in normal development
times (or sometimes ever). Some developers are selective in
a way that does affect overall contrast but the range is
limited, certainly less than one paper grade. Again, except
for adding bromide or Benzotriazole, there is little effect
on curve _shape_. BTW, neither variable contrast or graded
papers have consistent curve shape throughout their contrast
ranges. In most cases one grade different negatives can be
printed to have identical tone rendition by changing the
paper grade one step. The same with printing the same
negative on a condenser and a diffusion enlarger, but larger
variations may show some differences in mid-tone
reproduction.
  Contrast and image color. Image color of prints (and
negatives too) depends largely on the scattering of light in
the emulsion. This is in turn controlled by the particle
size of the silver making up the image. Extremely finely
devided silver looks bright yellow (its used as the filter
layer in Kodachrome). As it become coarser it shift toward
blue. Since the grain size of the image is somewhat due to
the developer very active developers that tend to produce
coarse grains also tend to produce bluer colored images.
Less active developers, for instance Kodak Selectol Soft and
Ansco 120 (nearly identical) tend to produce finer grains
and warmer images. They also tend toward lower contrast and
slower development.
  Tone Seperation.
   I am never quite sure what is meant by this. presumably
the ability to detect small variations in brightness. This
is partly affected by the resolution of the image. Blurry
images do not have good separation of anything including
tone values since they tend to blend into one another.
Partly, it is due to so called local contrast. At the
highlight and shadow ends of a print the contrast is lower
than in the mid gray area regardless of the paper, film, and
developers used. This is because the toe of the film is
affecting the shadow area contrast and the toe of the paper
is affecting the highlight area. Often, when photographers
complain of lack of tone separation in highlights (blocked
highlights) its because they are trying to print a greater
range of brightness than the paper is capable of
reproducing. The only solution to this is burning in, either
by hand or by means of a contrast mask of some sort. If
variable contrast paper is used the blocked areas can be
printed in using a lower contrast filter (with masking).
  The eye wants to see contrast similar to the original
scene so simply lowering the contrast of the reproduction
won't look right.
  As far as any developer claiming to modify the paper
curve shape (or film for that matter) the proof is in
properly done sensitometric testing. If there _is_ an effect
it will show up plainly on an H&D curve. There is a lot of
room for errors to creep in this sort of testing so one must
be wary tests which are not very carefully designed and
controlled.