Hmmm ... so you're not letting me duck the hard question, then. Sigh.

So your thought experiment is that we have an MS that looks as your
letterpress edition does, but we have evidence that the authorial
intent was to substitute the word "Theorists" with the word
"Speculatists", but then had second thoughts and decided not to?

This, in my mind, is not a substitution, but rather an aborted
substitution, for which the TEI has no dedicated element. I have to
admit that I haven't thought much about how to encode aborted
substitutions before, but off the top of my head I would be inclined
to either put status="aborted" (or some such) on <subst>, or (better
yet) status="intendedSubstitution" on the <add> w/o a <subst>.

Possibilities follow.

    <subst status="aborted">
This one is way weird, both because now <del> has a completely
different semantic when it's inside subst[@status="aborted"] and
because the rend= of the <add> (if I provided it) would be

    <del status="abortedSubstitution"><add status="intendedSubstitution">Speculatists</add></del>
Some might (appropriately) consider this stretching the meaning of
status=, which is for "the effect of the intervention", not the
intent of the intervention.

Better still, use the genetic encoding module, and describe what
happened in prose in <stageNote>s, perhaps indicating that the
addition was undone with an <undo> (or just <del>, or both, as I've
shown here):

          <ge:stageNote xml:id="sn01">
            <p>Word <mentioned>Speculatists</mentioned> inserted
              supralinearly, apparently in an aborted act of
              substituting <mentioned>Speculatists</mentioned> for
          <ge:stageNote xml:id="sn02">
            <p>The supralinear addition of
               <mentioned>Speculatists</mentioned> was deleted,
               allowing <mentioned>Theorists</mentioned> to stand as
               the end reading.</p>
  <!-- ... -->
       <ge:undo stage="#sn02" target="#add43"/>
       <del stage="#sn02"><add xml:id="add43" stage="#sn01">Speculatists</add></del> Theorists ...

> in principle, your suggestion is feasible, and may turn out to be
> the case if checked against the manuscript page. However, based on
> the rendering/typesetting conventions of this particular
> letterpress edition, I'd tend to disagree: horizontal alignment
> like this seems to consistently indicate a substitution;
> insertions, on the other hand, are consistently rendered in one of
> the following ways, depending on the presence and location of
> insertion marks in MS:
> * baseline caret and/or supralinear caret (carets can themselves be
>   struck out)
> * in the absence of any insertion marks in MS, the word is
>   horizontally positioned to indicate the logical insertion point
>   in the text
> So if it were an insertion, as you suggest below, based on the
> conventions of the letterpress typesetting, I would expect to see
> "Speculatists" positioned to the left of "Theorists" (assuming
> there was no insertion mark). (If this doesn't make sense in
> writing, I can provide an example from the print edition.)
> Thought experiment: For the sake of the argument, let's assume (1)
> that the example appears in MS exactly like on printed page, and
> that (2) we had strong supporting evidence that the authorial
> intent was in fact substitution, e.g. in the form of an authorial
> errata list or diary entry or what have you. Given those premises,
> how would you encode it?