A choral difficulty supporting the concept of harmonic functions

I held a rehearsal with our choir yesterday—Delibes, Messe brève, Gloria, in the SATB arrangement of AnnaMaria Hedin. Two alto singers—otherwise very capable and confident in finding (and leading others to) the right notes—did not hit a certain G, even when singing alone, without the "noise" of the other voices: Instead of G in m.139, they absolutely wanted to jump to the red E (or something like that, about a third lower than G). Not even the fact that G is just a semitone below the previous G♯, mirroring the semitone step from A to G♯, helped them:

For me, this supports the concept of a functional experience of music: The E major triad in mm.137+138 is the dominant of A minor; the G in m.139 is a (rudimentary) dominant of the—abruptly returning—key of C major. Thus, the alto first sings the third of the dominant of A minor; and then the root of the dominant of C major.

But if the singers do not consciously include the modulation in their mental model, they remain mentally in A minor and thus try to reach the root of this key's dominant! In other words, going from G# to G requires the higher mental effort of combining the elements modulate + go to root (+ remain in dominant function), whereas continuing with E can be reconstructed as (remain in same key +) go to root (+ remain in dominant function): The transition to E thus requires modification of only one state constituent instead of two, hence it is preferred.

Of course, this minimal experience does not say anything about the relationships or core set of functions (functional theory in its narrower meaning), and hence a Roman numeral analysis leads to the same result, with "dominant function" just replaced with "chord (or function) of the fifth degree".