Sure. I want to clarify that in Sean Graham's models and also the models that I've done for P.E.I. for the last six elections for the MP, pairing his neighbouring riding would be a convenient way to manage the data, but in practice it wouldn't be a prudent thing to have a boundary commission look at redrawing the boundaries rather than just pairing them because then you might end up with some quite strange riding shapes. But yes, you're correct. There's a single ballot, and we're using the primary candidate and the secondary candidate in every riding. To use Elizabeth's riding, if Elizabeth were elected as the primary candidate for an adjoined riding, then the secondary Green candidate who was not elected would be considered for the second seat in that riding, along with the first candidates from the other parties or any independents who ran in that district.
That is the list of candidates who have potential to be elected to the second seat in that district. The second seat allocation works by having the proportional calculation for the region as a whole calculated. The first seat's subtracted and which parties are due the remainder of the seats, then the aim is to award each party their deserved seats in the places where they have the strongest relative support.
In P.E.I. it would be a common situation when the NDP or the Greens did not win any of the first list seats, but they had their strongest performance in one particular area. It would try to allocate the NDP seats for their top candidates as the second candidate in that riding.