I'm thinking that is has something to do with the gluing technique.

Current state: If I understand your technique (and I may not - so correct me if I'm working on the wrong assumption), you first dry-fit the pieces. Once all of the pieces are positioned as you want them (square & true) them you apply glue to the intersections of the wood, effectively creating a fillet out of glue. A fillet's primary function is to create a mechanical restriction limiting lateral movement of the pieces of wood. Same as applying a hot glue to a joint. Very little or no absorption of glue into either piece of wood. Depending on the tightness of the fit some or very little glue will flow into the joint. IMHO there is no bond between the fibers of the two pieces of wood.

Others have shown that the strength of a glued joint is based on the amount of adhesive that is absorb into the two wood component's fibers. The two pieces of wood that your working with are lite ply of the former which has a fairly hard surface and a porous end. The keel is of balsa that should readily absorb glue through its sides as well as its end (but not an issue here).

AR glues aren't readily absorbed into the sides of ply but they do absorb into the sides of balsa. Maybe Titebond wicks into the lite ply (and its end) more readily than Elmer's glue. That may be why it held better.

If this were problem I would dip a cotton swab into the AR and cover the end grain of the notches in the formers with AR. Before it dried too long I would clean the notches with a wet paper towel or swab. This would allow the AR to absorb into the end grain of the lite ply leaving exposed dried glue. Others modelers have shown that surfacing wood end grain with AR provides an excellent glue anchor when the piece is finally glued with additional wet glue. I would mark the locations of the notches on the keel with a pencil. I'd pull or slide the keel away from the former and add wet AR inside the pencil marks, reposition the keel, and reclamp the keel & former at this joint. I would NOT try to do all of the joints at one time.

For a completely different approach that does not require repositioning after getting the former and keel aligned - how about using thin or maybe medium CA? CA wicks very quickly into the joint, up into end fibers as well as into the surface of the balsa keel.

pic 1 - glue as fillet only
pic 2 - glue absorbed into end as well as fillet