You're right; but actually the mention of 13 channels was just for the example. The algorithm works exactly in the same way for the 5G band as well (where you have similar problems as soon as you have a few neighbors using recent Wi-Fi devices).
Interesting, glanced through your paper. How do you find real devices behave at constrained channel widths (<20Mhz in 2.4, <40Mhz in 5)? I have a feeling those are not well tested scenarios :)
All 802.11n devices supporting the 5G band support at least a bandwidth 20 MHz (as well as 40 MHz). In addition, you can use 5 and 10 MHz as well on some Atheros devices with open source drivers.