The destination is acquired from the EQS and it will never be close to the start location which is in front of the AI pawn, it essentially finds a destination based on certain criteria and updates the blackboard, which the AI base class can access to retrieve a destination.
I then use the aforementioned function to generate a series of points that arc towards the destination and it generates the first point to start in front of the AI pawn. I later planned on turning these points to spline points to get a nice curve to travel along.
Your method is cleaner, but obviously I need to adapt it to automate how many points are required to face the destination. That is the only difference as far as I can tell.