Before you can answer this question you need to decide for yourself how the compiler should decide which expression belongs in which side of the AST. For example, should it parse:
a + b + c
exp1='a' '+' exp2='b + c'
exp1='a + b' '+' exp2='c'
Until you can decide this for yourself, there isn't a general case answer to your problem.
It may be the case that you are trying to put too much validation into the parser. Consider parsing 'EXPRESSION + EXPRESSION' in a single rule, rather than two, then deciding in your model validation and/or inference which version of the rule is intended (according to rules of your own choosing).