high-quality GAMESS-US basis set for iodine?

Having reviewed past CCL messages on the subject, I'm aware of the fact
 that the question has been asked before--but no good answer to be
 found...  Is it possible that there are none available?  The oft-cited
 PNL site offers a bunch of basis sets for iodine:
 STO-3G
 3-21G
 MIDI!
 WTBS
 Sadlej pVTZ
 LANL2DZ ECP
 SBKJC VDZ ECP
 CRENBL ECP
 Stuttgart RLC ECP
 DZVP (DFT Orbital)
 DeMon Coulomb Fitting
 DGauss A1 DFT Coulomb Fitting
 DGauss A1 DFT Exchange Fitting
 Ahlrichs Coulomb Fitting
 Huzinaga Polarization
 but, at least to my untrained eye, none of them seem to be of high
 quality, esp. if an all-electron basis set ready-made for use in
 GAMESS-US is sought.  The Sadlej pTZV set is not well optimized, at
 least based on my reading of the attached comments. I don't want to (or
 can't) use ECP and DFT bases, don't know anything about the last two.
 Why the problem?  Somebody was asking why [ICl2]- is linear, and I tried
 to use the GAMESS-US/MolDen duo to take a look at the orbitals, bonds
 etc., and noticed that the bond length at the 3-21G(d) level is wrong by
 0.07 A.  And then started looking for a better basis set, and...couldn't
 find it.  I realize the general picture at the HF/3-21G(d) level should
 be correct, but I'd love to have something closer to 6-311+G(d)
 basis--or better.
 What do the professionals use to get good geometries and energies of
 iodine compounds, even as simple as HI?  How much better can MIDI! be
 than 3-21G* in this situation? Is my aversion to ECP justified?  I
 appreciate any comments.
 I should note that doing such calculations are just a hobby to me to
 better understand the compounds I'm working with in an otherwise very
 practical millieu...
 Happy Y2K to everybody!
 Tony Gozdz
 agozdz # - at - # monmouth.com