Hi Martin
Good question. Modeling of collection types of various kinds (sets. ordred sets, bags, sequences, and arrays) is discussed in section 10.4 of the book. But the quick answer is that for asserted fact types, ORM currently requires at least a spanning uniqueness constraint. So the populations of asserted fact types are always sets of facts (no duplicates allowed). Currently we use a ternary as you suggest for modeling lists (a.k.a. sequences or ordered bags), For example, to store a list of prize winners for lotteries use "Lottery awarded prize of Rank to Person", where a UC spans the first 2 roles, and Rank {1..} indicates the prize order (1st prize, second prize etc.).
As discussed in section 10,4 of the book, there are some versions of fact-oriented modeling that do support modeling collections directly, but such structures must be used with great care and are awkward to popluate etc. We plan to have NORMA support a mapping choice so that you can choose to use such collection types at the implementation level, but we do not feel they are appropriate at the conceptual level.
Cheers
Terry