Tigramite stationary graph latent projection operation

[ghost]

Hi, while reading the 'tigramite_tutorial_general_causal_effect_analysis' file I came across the following section about the CausalEffects class:

"The above graph is a time series graph that is assumed stationary and, hence, repeats the above edges to the past and future. Internally, the CausalEffects class doesn't deal with these infinite graph, but rather constructs a finite ADMG with the same d-separations via a (more involved) latent projection operation for stationary graphs."

For my thesis I am using the Tigramite package and would like to write a clear description of this (more involved) latent projection operation for stationary graphs. I have tried looking in to the paper, but it did not clarify the question (at least for me) as well. Can anyone help to explain how this projection works, and what the reason is for doing it? Thanks a lot in advance!