[Raytrace] Optimization

[Richard F.L.R. Snashall]

I hope this is not too esoteric, as I will be describing a symptom
that I have encountered using some features in Zemax.  Starting
with a Clark refractor designed using the Gsum program (from Bob
May's website), I created an design (BK7,F4) in Zemax which I optimized
with the following goals:

      a) Effective Focal Length equal to 15 times the Entry Pupil
      b) Minimal Lens thickness equal to 11% of the semi-diameter.
      c) A default merit function of RMS wavefront error, not
         ignoring lateral color.

using an appropriate field and the built-in Photopic wavelength
selection:

        wavelength  weight
          0.470      0.091
          0.510      0.503
          0.555      1.000
          0.610      0.503
          0.650      0.107

After optimization, the Polychromatic Strehl ratio was 0.840
for that wavelength selection.  However, going back to ATM2,
and using the prescription given for the Baker version, the
Strehl ratio was 0.859.

I found (and I have found this consistent using several different
glass combinations) that, if I optimize instead for wavelengths
0.482, 0.522, 0.571, 0.628, and 0.670 (adding 12 ... 20 nm to
the each) that the Strehl ratio improves to 0.860.

What am I missing here, that the optimization using the desired
band is consistently coming up with a poorer solution for that
band than an optimization using another band?  Are these tools
measuring different quantities; quantities that are not strongly
correlated?

                        Rick S.