I somehow noticed
that in my idiolect, (1a) is much more natural than (1b), but (2a) is much less
natural than (2b).

(1a)    She is, after all, perfect.
(1b)    She, after all, is perfect.
(2a)    Nobody is, after all, perfect.
(2b)    Nobody, after all, is perfect.

I don’t know
immediately why this is. Nor did I know, until about 20 minutes ago, whether
this was an idiosyncrasy on my part, or a general feature of English speakers. Thanks
to Google, I now know that it is a general pattern
among English speakers. For a number of common pronouns and quantifiers, I
checked how often phrases like “She is after all” and “She after all is”
appeared in the Google search engine. The results
appear below. (“Before” means “after all” before the verb, as in “She after all
is”, after means “after all” after the verb, as in “She is after all”. Google ignores punctuation, so I didn’t have to worry about
things like commas. But that does mean that there are some false reports in the
‘after’ section, things where the match is to “… she is. After all, …”. I don’t
think this affects the conclusions I draw from the data.)

With the exception
of ‘someone’, and to a lesser extent ‘somebody’ there’s a notable difference
between the way referential phrases and quantificational phrases behave here. I
don’t know what causes the reversed distribution there, it could just be noise.

UPDATE: I forgot
to check for quantifiers involving ‘any’. Here are the numbers

Not much there of
note, except for the complete absence of sentences using anybody. Anybody,
after all, is entitled to use ‘after all’.