Is there an estimate for the typical number of morphemes in natural language?

The definition of a morpheme is that it is the smallest meaningful unit of language. Morphemes do constitute words but they also include root words. Therefore there's no point in trying to enumerate them in a single language let alone across languages.

You could enumerate things like inflectional or derivational morphemes which would are always very limited inventories but that would not really help you.

The point of the double articulation thesis is that you have a system of articulation meaning (morphemes, words, sentences, texts) that is built upon a system of articulation without meaning (phonemes, syllables, etc.). It is not intended to support a separate thesis about the productivity of language associated with Chomsky.

The thing to remember is that although strictly speaking morphemes are built out of phonemes, phonemes actually don't combine into morphemes. They combine into syllables which do not have a straightforward relationship with morphemes (a morpheme can consist of two syllables, e.g. 'simple' but one syllable can also represent two morphemes, e.g. 'runs'). Syllables then combine into metric/intonational units which again do not always map straightforwardly onto words, phrases, clauses and sentences.