The Lunar Orbiter probe series collected clips of analog video, which were stitched together on Earth to make maps. This was done by hand in the 60's but later the video was digitized and newer maps were made by machine processes in the Lunar Orbiter Image Recovery Project from 2007 to 2014.
The initial maps were not high quality, because of changing perspective and a wide field of view images had to be distorted to stich them together, and nothing worked over large areas, the inaccuracies would compound with the curvature of the surface. Local maps for Apollo landings were hand drawn from the source images, but a full globe wasn't produced until much later and when it was the LOIRP process only makes up for so much.
What you're seeing here is a stitching issue where neighboring images couldn't be reconciled - either there was a gap without coverage or the only adjoining images had perspectives too far apart to reconcile into a vertical view. So the result, you see a crater and nearby features split in half with straight lines connecting the halves, as well as distortion to surrounding features (particularly noticeable in the two craters bordering the left side of the red circle.
Clementine launched thirty years after the Lunar Orbiter program and had vastly superior imaging, and had additional tools specifically to allow accurate global mapping without the issues of earlier maps. It had a narrower field camera with higher resolution, but also had additional ultraviolet, near-infrared, and laser instruments. The Google Moon Clementine globe was machine produced from that data in much the way the LOIRP map was produced from the older set, but because Clementine's instruments were more advanced and more numerous, they provided vastly more data to use in reconciling issues, so they include fewer errors (and because the craft wasn't in exactly the same orbit those errors are not in the same places as those of the LOIRP map). The areas inside the crater is something peculiar to processed Clementine images. Clementine's mission didn't last that long, compared to the Lunar Reconnaissance Orbiter, so there's not a lot of redundancy in its data set, often no second pass on a feature with different lighting to fill in details. The other sensors can be used to approximate details in those cases.