It indeed looks like a confusion. Energy levels characterize some material system (electronic levels of Atoms; Electronic rotational and vibrational levels of molecules; electronic bands or phonons in solids and many other material systems). Electromagnetic waves exist with no need of material. The thing is, that electromagnetic wave might interact with material. The interaction is particularly strong if the frequency of electromagnetic wave fits the characteristic frequency of the material system (some of its degrees of freedom). THz frequency range might fit typical frequencies of vibratons of rather heavy molecules, also it might meet some rotational spectra. Such studies have a lot of sense, and there is a whole branch of THz spectroscopy. The THz generation mechanism as you described might, in principal be described in this quantum language, if we will look at the electronic energy levels of nitrogen molecules close to dissociation. The problem, nevertheless, is that close to dissociation, the electronic energy levels become so close, that they merge, so we effectively get a continuum of energy states. Basically it means that electron can have any energy within an allowed energy band (range), so, in such a situation, it becomes much more intuitive to use classical physics of electrons moving in electromagnetic field, rather than keep quantum-mechanical formalism.

After all: any type of theory is just a toolbox of mathematical tricks to explain a behavior of physical objects. Classical physics in this case is simply easier to use, and in most cases gives similar results to a more complicated quantum picture.