I think that you've asked a complicated question. I'll try to split my answer into several parts.
First, on the topics of "duration" and quantization. A mora is a phonological unit. As with vowel quality, "length" is an area where languages appear to construct phonological categories that are discrete even though phonetic duration is naturally gradient. (Gradiency might also exist in phonology in addition to discrete categories, but phonological gradiency is a more controversial concept that you don't need to investigate to understand the concept of a mora.) A mora does not directly equal a unit of phonetic duration (although pedagogical descriptions of how moras function in languages like Japanese often simplify the idea by describing it in terms of phonetic duration).
As melissa_boiko said in a comment, it's typical for long syllables to not be realized with twice the phonetic duration of short syllables. Given that, you might ask why McCawley specified that long syllables contain "two" moras? It's not just because two is the next integer after one. In some languages, we can find evidence that heavy syllables are in some way equivalent on the phonological level to two light syllables when we look at the structure of some kinds of poetry. For example, Japanese haiku is typically described as being based on mora count, and Classical Latin quantitative meter allows the option of either two short syllables or one long syllable in certain positions. However, such evidence does not exist in all languages, so the situation is not always clear. The occurrence of moras in English for example is more debatable.
Some languages have been analyzed as having trimoraic syllables based on the “extra-heavy” behavior of syllables ending in VːC/VVC or VCC, but I’m not sure what kind of evidence has been found indicating that superheavy syllables contain specifically three units of length. (In Classical Latin, there are syllables ending in VCC or VːC/VVC, but they behave identically in terms of prosody to syllables ending in Vː/VV or VC, so I see no basis for saying that Classical Latin syllables can be any longer than two moras.) Since you posted the question, I have started looking at some papers by Kevin M. Ryan that have some interesting discussion of the topic of phonological weight. In a 2016 preprint of the article "Phonological weight", Ryan cites Bruce Hayes 1979, "The rhythmic structure of Persian verse", for the statement that in some kinds of Persian verse, superheavy syllables are treated as metrically equivalent to heavy + light sequences. In that case, it seems like the math does indeed work out if superheavy syllables are analyzed as containing three moras (vs. a regular heavy syllable containing two moras), since 3 = 2 + 1. However, I haven't been able to get my hands on the 1970 Hayes paper yet, so I can't describe the relevant Persian data any further.