I've never used a Faro Focus but I do own a Leica C10 ScanStation and I imagine that the process of merging point clouds is similar for all terrestrial laser scanners. You generally need to have three or more tie points between pairs of point clouds captured, with each point cloud derived from a single scan station. In the case of my system, there are specific targets that are used for the GCPs. You set these GCPs up within the scan field of a station, positioning them such that they will also be visible from the second station. In fact, if you are able to locate them such that they are visible from multiple scan stations, you will need to do far less work in the field. Importantly, you do not move the targets between the two (or more) scans, i.e. only the station moves. That means that you need to have at least three targets with you when you are surveying. After completing the scan, you then specifically zoom into each of the 3+ targets and perform a small and intensive scan of the target. These are usually saved on the system as a separate tie point data set.
Then in post-processing with the software distributed with the system (Cyclone in the case of the C10), these data are used to combine your multiple point clouds. If you also have GPS coordinates of either the scan stations or the target locations, you can also georeference the point cloud. Importantly, this requires some planning ahead of time (bringing the appropriate targets to the field) and during the scan (positioning the targets in highly visible sites within). In my experience, this aspect of positioning good target locations and scanning them from each station is the most time consuming part of creating a multi-scan data set. With my system, just locating the targets in the onboard camera/screen on a bright day can take a substantial amount of time.
If you do not have targets within your point clouds, some software will let you pick common tie points based on features mapped within overlapping clouds. This is not the most accurate way to carry out this type of survey and is dependent upon the presence of recognizable objects/features within pairs of point clouds. This is generally more achievable in urban settings (e.g. using building corners, tops of lamp posts, etc.) than forested sites. The accuracy of the merging will depend on how well positioned these tie points are, how many there are, and how they are distributed throughout the data set. When people talk about the accuracy and precision of terrestrial LiDAR systems, they are usually quoting the system specifications, which are normally stated for single scans. People often overlook the fact that the process of merging multiple point clouds introduces a different and significant source of error in the final data product.