In CGEL on pp. 1316-1317, we find a discussion of two different uses of as well as:
[70] i b. [Abstraction] [as well as impressionism] were Russian inventions.
ii a. [Beauty] [as well as love] is redemptive.
(The emphasis in boldface is mine.) The analysis goes like this:
In [i] as well as behaves like the coordinator and. ...In [ib] the form were indicates that the subject NP is plural, just like abstraction and impressionism. ...
In [70ii], by contrast, as well as behaves markedly differently from a coordinator. In [iia] the 3rd person singular verb-form is indicates that this time the subject is singular: is agrees with beauty, so that as well as love is treated syntactically as an adjunct, not a coordinate. ...
We must conclude that idiomatic as well as can be construed syntactically in two ways, introducing an element that is either coordinate (as in [70i]) or subordinate (as in [ii]). In the former case, we take it to have been reanalysed as a compound coordinator. In the latter case there has been no such syntactic reanalysis, and here as well as does not form a constituent. This is evident from the fact that as well can occur on its own: compare Beauty is redemptive and love is as well. In [iia], then, the second as is a preposition taking the NP love as its complement, and the whole PP as love is an indirect complement in the AdvP as well as love.
My question concerns how this argument makes use of the sentence
[B] Beauty is redemptive and love is as well.
This seems to be a nontrivial rewriting of [70iia]. What justification do we have for the claim that the as well in [70iia] and the one in [B] are 'the same' as well?
The 'data' show just this: that the idiomatic as well as seems to function in at least two different ways, and the way it functions in [70iia] is one in which it is not equivalent to a coordinator. CGEL says that the difference between the occurrences of as well as in [70ib] and [70iia] is that as well as is a constituent in [70ib] but not in [70iia]. And CGEL's evidence for this assertion is that as well appears alone in [B]. Thus, CGEL's argument assumes, rather than proves, that the as well in [70ib] is the same as the one in [B]. So: how do we know that this assumption of CGEL's is true?
To repeat, all we know is that the as well in [70ib] is not like the as well in [70iia]; but why does this imply that therefore the as well in [70ib] must be like the as well in [B]?