I am trying to understand the shape of the dotted curve in the SED plots.
Did you take a Starburst99 model and then ran it through some dust
attenuation prescription to obtain the shape? In this case, the UV part of
the dotted curve would be mostly a reflection of the attenuation law. Is
this correct?

--I did not apply any attenuation law to the curve.  Looking at St99
  website plots, I see that the 900 Myr spectrum I utilize is more like
  the instantaneous SF version rather than the continuous SF version.

  Below is the input file I used for retrieving the St99 curve.  I see
  that I used ISF=-1, which means I should be saying 'instantaneous'
  instead of 'continuous' in my fig caption.  

A 900 Myr old burst is not an appropriate comparison. Basically you are
saying there are no stars with masses above about 3 M and that there is no
stellar UV radiation. What should power the dust in this case? I realize
that nothing quantitatively has been done with the stellar SED. If this is
the case, why show it at all if it is wrong and misleading? The continuous
SED would be roughly flat between 0.1 and 1 �m. One could argue that the
difference between the observations and the model SED indicates internal
dust attenuation. Right now, the majority of the UV data points are above
the model SED, leaving no room for attenuation.

--Thanks for the feedback, Claus.  Various reviewers liked the idea of a
  standard template as a reference from galaxy to galaxy.  If we are to
  keep this idea, which type of template should I use?

Obviously the stellar populations in an elliptical and a spiral galaxy
differ. Therefore they do not share the same template. However, our sample
is probably closer to late- than to early-type galaxies. The fact that we
see dust suggests at least some relatively massive stars. Therefore I would
choose a template with continuous star formation. The 900 Myr example you
describe in the text is actually a good one. You had the right model in mind
but you just clicked the wrong button..:-)

--Done

BTW, Vazquez and Leitherer 2005 use a Croup, not a Salpeter IMF. Croup is
the default on the website. What actually did you use?

--Well, I did use 
  XPONENT=2.350
  when I inadvertently grabbed an instaneous model.  But today when I
  reran the sim (to grab a continuous model) I used
  XPONENT=1.300,2.300
  since you use that as the default.