Q: Why does Globasa blend words à la Lojban, particularly Sinitic words? I think it makes more sense to pick one form that directly resembles one language and is still recognizable to speakers of other languages than a Lojbanesque mashing of words that will be vaguely recognizable to many people, but not directly to anyone.

What you propose is certainly one valid way of going about it, but it simply isn't in the spirit of Globasa, as you'll see below.

But first, in reality, the comparison with Lojban is unwarranted. Take a look for example at the word for "house" in Lojban: zdani. Can you tell me what the etymology for this word is? What natlang words is "zdani" similar to? None.

Yes, Globasa sometimes blends words together out of multiple languages, and this is often the case with Sinitic words due to their often great variation in form, at least greater than that of European languages, for example. (By the way, in European languages, there's a greater incentive to keep forms intact, and that is spelling recognizability.) At any rate, in spite of that, Globasa doesn't come close to Lojban's word blending system. In Globasa, the idea is to blend a word in such a way that it still very closely resembles at least one of its source words. Usually there's a difference of only one phoneme (plus/minus one phoneme, or a different phoneme, often changing only one articulation feature in the case of consonants: place, manner or voicing).

So Globasa's approach is actually not too far off from what you're suggesting. It's certainly closer to your preferred approach than it is to the system in Lojban. For example, funsu (fountain) is very close to the Japanese "funsui", ijen is very close to the Japanese "iken" as well as to the Mandarin “yìjiàn”, etc. Yes, sometimes there will be a difference in two phonemes, but if so, the second consonant typically only changes one out of the three features.

Why not just go all the way, and instead select "iken" or "ijyen", for example, so that we end up with the most "naturalistic" words possible, and so that the word is directly recognizable to at least some people? Well, because it just wouldn't be in the spirit of Globasa. One of the main principles in Globasa is to strive for a middle ground, which is a pragmatic approach that attempts to spread out the burden of learning.

Also, blending words is consistent with Globasa's guiding vision (the island thought-experiment). Look at it this way. When we say "naturalistic", what does that mean? Typically, that's taken to mean what you suggest, or what Lidepla does. However, in relation to the island thought-experiment, a naturalistic outcome wouldn't be that natlang forms would necessarily remain intact. In a natural process of creolization, forms would tend to morph, and so from that point of view, selecting intact natlang forms is actually rather artificial.

One final reason for allowing blended forms is that in this way we can more easily avoid certain minimal pairs, since we have even greater selective power when determining word forms. This could even sometimes result in a natlang word remaining "intact" in order to avoid a problematic minimal pair.

By the way, in some cases "blended words" happen to coincide in form with the word in a language of the same family as the primary source languages. For example, the word salom, which is blended from Arabic "salam" and Hebrew "shalom" is identical to the word with the same meaning in Uzbek.