This is just that one author's opinion - it is telling that both sources you have for the number come from the same author. Henry Markram has a leadership role in the Blue Brain Project, a project to create a very detailed computer simulation of a chunk of neocortex to answer various questions. In the context of this project, it is somewhat necessary to make some decisions about numbers of cell types so that they can be incorporated in the model. People can bicker all day long about which types are actually unique types, though.
As far as the mathematics of having 55 morphological types and 11 electrical types, yet only 207 morpho-electrical types rather than 605, the answer is that the matrix is sparse. To get 605 types, you would need to observe all 11 electrical types within each of the 55 morphological categories. If not every electrical type is observed for a given morphological type, you will have less than 605 total. In some cases morphological types are specifically associated with particular electrical types.
You should also read about the concept of lumpers and splitters - in summary, in any context of classification without clear distinctions, different people will come up with different boundaries. It isn't really possible to argue "right" versus "wrong" in this context, because doing so requires weighting different values. Lumping may be beneficial for the sake of simplicity and generalization but miss some details; splitting may allow for more nuance but also risks not seeing "the forest for the trees."