I meant the bonus question for Daivid, but well… 
In theory, Greek words can be divided into morae, short vowels constituting one mora and long vowels and diphthongs two. Whatever the true nature of the accent, it’s convenient to assume that it is really a single mora that carries the accent and not the whole syllable; the mora with the accent has a rise in tone and the next mora has always a fall in tone - two consecutive morae cannot have the accent, because the tone must always drop to baselevel in between.
If a syllable is long (i.e., has two morae) and has the accent, the accent can be either on the first mora or the second. If it’s on the first mora, it’s a circumflex (e.g. ἐγεῖραι egEerai, the majuscule E representing a rise in tone on the first mora of ει, the following minuscule e a fall in tone on the second mora of ει); if it’s on the second mora, the syllable is an acute (ἐγείραι egeErai). If the syllable is short, it has only one mora, so the accent can only be acute (ἔγειραι Egeerai).
λίπέ με is impossible, because it would require a rise in tone in two consecutive morae (*lIpE me). With a circumflex, the tone falls in the second mora of the syllable, so the next syllable can be acute (σῶσόν με sOosOn me).
(I’m aware of one exception to this rule: with interrogative pronouns, there can be an acute on two succesive morae: τίς τ᾽ ἄρ σφωε θεῶν ἔριδι ξυνέηκε μάχεσθαι; (Il. 1. 8.))