It seems like the low-frequency inter-modulations are responsible for the "perception" of ultrasonics. This, together with frequency aliasing due to improper low-pass filtering may probably explain the perceptual differences between the recordings made with different sampling rates. I don't think that these issue may have affected the development of our upper frequency limit - we had no loudspeakers in the forest/jungle :-)
I like the rephrasing of my question like Kevin did. Thinking in wavelengths directly points to sound localization, which of course, relies on the ability to detect sounds. Maybe our sound sensitivity between 8 and 16 kHz just followed the development of the head size - just because this was the range providing large ILDs and humans with lower frequency limit simple could not localize the sounds that well. What if the development of our pinna followed the sensitivity in that frequency range? I wonder if a species having our head size but being deaf beyond 8 kHz would develop pinnas with doubled size...
By rephrasing my question, now it sounds quite speech-unrelated. Nevertheless, an interesting aspect remains: how is the speech related to our upper frequency limit? If a species were able to detect (and localize) sounds before it developed vocalization, speech would have no contribution to the development of the frequency limit, right? Frequency range of the ancestors of homo sapiens would be interesting, especially related to the analysis of the FOXP2-gene, which seems to be responsible for the ability to develop vocalization. Any references? ;-)