The underlying formalism for computing the transmission coefficients T(E,k) for a particular energy E and k-point k is actually to take the trace of [tex]t^\dag t[/tex] where [tex]t[/tex] is the transmission matrix which you can view as the coupling matrix between incoming states in the left electrode and outgoing states in the right electrode ([tex]t^\dag[/tex] is thus the opposite).
However, we can also diagonalize the matrix [tex]t^\dag t[/tex], in which case we will get the transmission eigenvalues you are referring to. These basically tell you how many channels that contribute to the transmission, and their magnitude. Because of the invariance of the trace under unitary transformations, the sum of the transmission eigenvalues equals the trace of the original matrix, i.e. the total transmission coefficient T(E,k).
The reason the older code reported 3 values was simply that it was hardcoded to show the 3 largest eigenvalues. The newer ATK will compute and show all non-zero eigenvalues.
