Apple GCR 5x3 nibblizing question

[eric@rothfus.com (Eric J. Rothfus)]

Anyone out there know the details of nibblizing
for the very early Apple 5x3 GCR?  There's
a description in "Beneath Apple Dos" (BAD) but it is
rather weak, and not very precise.

I'm looking for the details of the nibblizing
process that separate the bits from the target
256 bytes into the 410 "bytes" that will then
be mapped to GCR5x3.

The BAD (no pun intended) description of the
process indicates that the 256-byte source data
is split into 5 sections


   0 -  51 - section "A" bytes (52 total bytes)
  52 - 102 - section "B" bytes (51)
 103 - 153 - section "C" bytes (51)
 154 - 204 - section "D" bytes (51)
 205 - 256 - section "E" bytes (52)

This is taken from the diagram on 3-16 of BAD.
Note that there appears to be one extra byte
in this range according to the diagram.

One guess is that this is off by one in
the E section, changing it to:

 205 - 255 - section "E" bytes (51)

However, if this is the case, there are too
many bytes in the A section such that the
"secondary" buffer wouldn't balance when constructing
the bytes made up of 3 A bits, and one bit from
the D's and E's.

Another guess is that the size of the sections
are all 51 bytes, with one byte left over.  This
would make everything balance well.  EXCEPT that
there is still one byte left over (original byte
255) that would have to be nibblized and stashed
somewhere in the secondary buffer in 2 nibblized
bytes.

Anyone know anything about this?  The 6x2 description
in BAD was pretty poor and not terribly accurate
either.  In the second addition of BAD, BAPD, 6x2
is described correctly, but 5x3 is not covered.

Thanks,
Eric