I would have all these frequencies were in range, but when I search I find several sources suggesting that 600 and 900 Hz are indeed greater than the fundamental frequencies of human voices:
Baby cries have a fundamental frequency (hereafter referred to as Fo) of around 500 Hz (roughly corresponding to the note B4). Child speech ranges from 250-400 Hz (notes B3 to G4, adult females tend to speak at around 200 Hz on average (about G3), and adult males around 125 Hz (or, B2).
My guess at this point would be that while the "fundamental frequencies" are lower for vowels, higher frequencies are still required for the consonants?
Now - maybe humans create some higher frequencies, but the dominant frequency you speak at is much lower.
Maybe I can understand some 'peaks' at higher ranges ... but the 'big peak' should be at the frequency that is obviously the loudest, which is 100-200Hz. That's the range you speak at.
But I also read that those are in fact the 'same frequencies' - simply at different phases.
I also read that without 'windowing' (apparently an extremely complicated process) you get wierd results.
I did a test with a 'pure' 300Hz tone - and I got a nice peak at 300Hz, but some other, rounded peaks at other ranges.
Anyhow - I don't think I'm doing something wrong - I do think that FFT in the real world definitely has some weird artifacts that require interpretation, and they are non-obvious.
