Welcome back, R-Statistician!
(Even though only your debut post precedes this,
and was 4 years prior, you are remembered.
)
I do not have access to the Pawson(-named) document
that you cite, but do have two earlier works under his
name. I mention "name" in consideration that, for
whatever reason, there might be separation from the
contribution of the author with the illustrations in the
book --often to poor result, or otherwise (not so) surprising
result. ("Don't these guys even TIE their knots?!" I'm
inclined to wonder, at times.)
Both
The Handbook of Knots &
Pocket Guide to Knots & Splicesshow the (more) secure form of the
carrick bend. The
first(-published) is not explicit about the position of the
tails; it contains the curious instruction to "pull on all
four ends to tighten the knot" --huh? (eariler it says
that the knot can collapse when loaded). The latter
work expressly advises that tails should be "opposite
on another and not too short, as the [end-2-end knot]
will collapse when put under strain." This is frightenly
shy of decent knot advice : one SETS the knot into
form, preferably, not hoping for favorable capsizing!
(And, as has been presented in the IGKT's
Knotting
Matters (by P.vdGriend) and briefly filmed on the t.v.
show
"The Deadliest Catch" (to my eye), the knot is tied
by another method that doesn't entail capsizing.)
As for the
stevedore knot, that is shown correctly in the
former work and isn't in the latter. I surmise that what
you describe as the mistake is a like knot with one fewer
(half-)turn --what rockclimbers might call a
"fig.9" knot;
this is a not uncommon mistake seen in knots books.
--dl*
====
ps : Hoping that you're beyond that old (back then!) computer
system, and so able to help to keep our musings tamed in
statistical implications.