You can call functions with Execute Function action task on a gameobject monobehaviour even if you creating asset-based behaviour trees and are dynamically adding the behaviour tree in runtime, as long as the gameobject does have the monobehaviour that the Execute Function targets when it executes of course.
The same is true for all reflected tasks. You can just select the target function/field/property of a monobehaviour, even when you are designing asset-based graphs. In runtime the task will work if the monobehaviour and target member (function/field/property) can be found on the target gameobject.
Please let me know if that helps or if you need further clarification.