It figures, as soon as I posted my question, I heard back from Tracy:

Hi Dave,

That is as puzzling to me too! My first reaction is that it should say
Z = Y1 + (Y0.bit15)
the purpose being to round off the value to output. That should be done based on the
high bit of the low word of the accumulation. It is not usually going to be
important to round it off. With Z = Y1, it is always rounding down.

-- best regards,
Tracy