This is what I believed when I created them. In the case of Megdevi, it came directly from (a) learning about affricates, and (b) realizing that Ancient Greek had the letters ξ and ψ. I assumed they were affricates and phonemes, because they have single letters for them, but they’re not. Technically an affricate has to feature a stop and fricative in the same place of articulation (with some wiggle room allowed on the symbol used for either element. For example, the /t/ part in English /tʃ/ is not in the same place of articulation as our actual /t/). Even so, it seems trivial to take aspirated stops and turn them into stop + [s]—i.e. *pʰ, *tʰ, *kʰ > ps, ts, ks. I’d be surprised if that didn’t happen somewhere, even though actual affricates is more likely (i.e. pf, ts, kx). At that point, they’d be acting like single consonants, because that’s what they were. Still, I think what they’re trying to capture with the idea of the affricate is that it’s a stop with a crazy slow release—so slow that your tongue (or lips) hold on a fricative sound before you let the whole thing go. For that reason, something like Megdevi’s [ps] and [gz] couldn’t be affricates, because no matter how slow you release a [g], you’re not getting to [z]: that takes effort on the part of the tongue.

That said, there are phonemes that begin as one thing and end as another that aren’t affricates. For example, prenasalized stops like [ᵐb], [ⁿd], and [ᵑg], or labiovelar consonants like [kp], [gb], and [ŋm].

You should read what I wrote here on Megdevi before calling it one of the best conlangs. There are many mistakes there—including mixing up clusters and affricates. The following can be affricates (can also be sequences): /ts/, /dz/, /tʃ/, /dʒ/. The rest of the ones in Megdevi are not. I was led astray by Greek, which has separate letters for the sequences /ks/ and /ps/. Just because it has a letter, though, doesn’t mean it’s a phoneme! (So, of course, it’s still nominally all right in Megdevi, if you’re just looking at the orthography. Unfortunately the other ones actually act like phonemes, so it’s just a mistake.)

Kind of, but it wasn’t unintentional. When I created my first language, it intentionally had the regularity of Esperanto (a language I was studying at the time) and a triconsonantal root system similar to Arabic (another language I was studying at the time). In some ways, that first language was very similar to Arabic and Esperanto; in other ways, very different. You can judge for yourself here. I never had enough familiarity with a conlang beforehand to create something that was unintentionally similar. It was a different era. When I started, I didn’t know of any other conlangs than Esperanto and some of its competitors (Volapük, Ido, Novial, etc.). I didn’t know there were any other living language creators, either. It wasn’t until I was further along that I found other language creators, and by then my preferences were set, even though I had a lot to learn about the nuts and bolts of language creation.