As Julian Dyer has observed, the roll manufacturing process could vary,
even with the same roll number.  If you want to see this in spades try
scanning nickelodeon rolls!  One must consider the target market for
nickelodeon rolls to understand why the manufacturers weren't very
careful, to put it politely.  I have seen rolls that had step size
variations within a tune up to 5%.  The first time I saw this I thought
my scanning system was at fault, but careful measurement of chain steps
showed the problem to be in the roll.  All due credit to the software
for sticking the punches back on even steps.

It's a head-scratcher why step sizes varied so much even with the
same manufacturer.  I thought it might be due to paper shrinkage or
expansion, but careful measurement of the roll widths showed this to
rarely be the case.  Another possibility is that instead of direct gear
drive they used some sort of continuously variable system to set the
step size.  Considering that the variations are often less than a
percent, this may well be the answer.  Does one copy the step sizes
exactly or try to figure out what the original intent was?  Since my
process allows for punch step size resolution in the order of three
thousandths of an inch I have opted for copying the step sizes of the
original rolls.

Today's scanning processes are quite good, but can be fooled by such
things as poor punches, translucent defects in the paper, skewing of
the punches, as well as repaired damage.  I have found that the only
way to be sure that the scan is correct is to cut a copy roll and
overlay it with the original.  Even this can be imperfect, as the
comparison is only as good as one's eye, and it's easy to miss a single
step error, particularly with the finer step sizes found in reproducing
rolls.

Wayne Stahnke created software for comparing two scan files, and his
method was to scan a roll several times and verify that the files were
the same.  This could be done with a scan of the original roll and one
of the copy, but some errors will be faithfully reproduced in both scans.
It would be possible to scan the original roll in both reflective and
transmissive modes and compare the files, which should eliminate many
of the scanning errors.  So far I haven't found a better way to
approach accurate copying than tedious manual comparison.

It bothers my perfectionist thinking to realize that we are not going
to get exact scans or copies of rolls.  Then I think back to something
one of my bosses said to me long ago, "We'd be out of business if we
tried to be perfect.  Do the job 'Good enough'."  That can be close
enough to perfection that most of us humans will never know the
difference.

Bob Billings