
Only the "final" vectors have been constructed with the correct
zeropoint so that H0=65 (using the Hubble flow sample specified 
in JRK06 and the rest-frame fitter). 

For the other vectors the normalization of the distance modulus
will not be exact (but it is pretty close...)

vectors:

final     -10d to +90d MLCS2k2 vectors with proper training
early     same as "final", but extended (by hand) to -20d
tweaked   similar to "early", but templates shifted up/down 
            slightly, to better match color/extinction zeropt
oldadam   vectors/Smatrix in Adam's homebrew code

rv19-orig -10d to +90d with R_V=1.9 training
rv19-early extended by hand to -20d

S matrix options:

sorig      raw S matrix from the training set
slowz      same as orig, but all errors scaled up by ~2, to
             come closer to low-z dispersion, used in JRK06
             with the rest-frame fitter
slowznoff  same as slowz, but off-diagonal elements set to 0
s=0        S matrix is 0 (use only for testing purposes)
