It's rather important to understand why you need to set up the transport model with more than just a few CNT layers.
To begin with, computing the transmission spectrum of a perfect CNT - or for that matter any perfect 1D system - is rather pointless. All you will get is T(E) = integer values = the number of bands at each energy. So you just need the band structure to draw the transmission spectrum.
Even so, it's of course a rather good system to start with, to understand some basics.
First of all, there is the issue of how long the electrodes should be. This is discussed at length in
http://quantumwise.com/publications/tutorials/mini-tutorials/99; for a metallic (n,n) CNT with a period of about 2.46 Å, the outcome of this analysis shows you need 2 periods, and ideally 3 to avoid artifacts which would result in precisely the kind of dips you see (for DFT; for Cerda Huckel parameters, you'd better have at minimum 4).
Given 3 periods in each electrode, this means we need minimum 6 periods in the center.
An alternative to computing the perfect setup it as a transport system is to compute the transmission spectrum directly from the bulk configuration of the nanotube, with 3 periods (i.e. corresponding to the electrode). This is a much faster calculation, obviously, but of course it only works for the perfect system. And, as mentioned above, you could just do the band structure and count lines, however that approach only works for 1D systems; the T(E) from bulk approach works for 2D and 3D systems as well.
It can be mentioned that the dips are typically only seen for perfect systems - or only detected for them, anyway. And in one way they are not crucially important (except for the "niceness" of the plot) - if you want to compute the current, small single-point dips integrate to nothing. Once scattering is introduced, such small details in the transmission spectrum can however be indistinguishable from real features in the spectrum. So, in a sense testing the perfect is a good way to determine the effects of the electrode length - but doing the analysis in the mini-tutorial is really much quicker.
