Exception 6

PE19/58

But in normal long words, especially those made with derivative suffixes etc., the secondary accent was attracted by a long syllable (not adjacent to stress), and never fell on a short final syllable. A strong secondary accent was borne by a long syllable separated from the main (a) by one syllable, long or short: types màndălḗ, kártastā̀; or (b) by two syllables: type kárpalimā̀. A weak secondary was borne by the furthest of a series of two or three short syllables that preceded the main, types àlakánda, kằlŭmălánda, kằlŭmălṓ.

Note that a long syllable adjacent to the stress did not attract the secondary, so èrĕmăloitḗ.

Note that this paragraph talks about CE stress before the Quenya stress shift. However, we assume that the logic of secondary stress placement stays the same:

PE19/60

The secondary stresses were usually reorganized and governed by the main stress; but on same principles as those described above for the archaic period — except that a secondary accent (often[?] with low tone) could stand on an initial syllable immediately before a long main stress, as in cṑmallo above.