Hi:
I don't feel there's enough information to even guess at an answer. For example...you have a dataset with ONLY 1 variable or column (COL1) and some number of rows or observations?? There are no identifying variables?? It's also sort of hard to "forget" that you mentioned the possibility of a variable number of rows to add up (maybe 3, maybe 12). You do not want to use LAG, but why not?? Are you rejecting LAG for some reason??
Possibly transposing the data and then using an ARRAY might work. Really, without more information, it's like putting a 500 piece puzzle together with someone who's handing me only 3 puzzle pieces at a time.
Perhaps other folks will have some better ideas.
cynthia