I can’t tell how much vertical discrepancy there is between the two clouds, but if it’s obviously out of tolerance I would first suspect processing methodology and/or control point values. Even different runs should be matching up in the overlap areas.
Some of our larger projects incorporate both mobile and aerial LiDAR, and the mobile mapping runs usually match up with our sUAS point clouds within 1-2cm, using different processing software, different raw data and often slightly different methodology.
Sometimes the data are gathered weeks or months apart. It’s all about the control and the processing workflow.
As for fixes, cloud to cloud is going to be tricky with that small amount of overlap, especially if there aren’t many identifiable features in the common areas. I wouldn’t have much confidence in a river bed to extract common points in automated routines.
“…people will come to love their oppression, to adore the technologies that undo their capacities to think.” -Neil Postman