We improved identity resolution with clearer ranking logic and stronger edge-case handling.

For common names, we now require multiple confirming signals (such as location and gift history) before assigning a higher confidence.

For uncommon names, matches rank higher by default because of lower ambiguity.