The Besselian to Julian Epoch conversion routine may require additional clarification on inputs

I am using the Besselian Epoch to Julian Epoch conversion routine to translate NASA Shuttle Position data from M50 coordinates to J2000. The documentation on these routines simply states that the Position and Velocity vectors should be provided as input. Providing the two 3x1 vectors results in a good conversion for position but a bad conversion for velocity. I get reasonable velocity conversions when I unitize the position vector first.

1) is this expected behavior, and 2) if it is, I recommend clarification in the documentation.

Of course if I have mis-interpreted the documentation, please set me straight. I have enclosed a couple of screen captures illustrating the issue.