Hello Alikhan,
After Align, a new Component(s) forms containing a sparse point cloud, but only after clicking Preview, Normal, or High does your Component now feature tie points, these being selected from features that were detected during Align, but then were found as common features between overlapping photos or cameras. What were simply 2D features now have a Z value for their depth. If you enable “Tie Points” under the SCENE tab, the Lasso tool is made active under RECONSTRUCTION, you can then select unwanted points to then apply Filter Selection to delete them (forms a new Component), or — most useful — you can rotate the model to identify separations in a wall due to misalignments> click Find Images to locate the problem children cameras> disable them> run align to form a new Component without those cameras. A this stage RC doesn’t have all the cameras you want active, but at least your Component won’t contain that particular group of misaligned cameras, which only causes problems with control points showing a red exclamation point (high reprojection error (when you’re confident your CPs are good), also the problem of inviting more misalignments downstream when bringing in other photos, gets too complicated to parse this out how the problem ripples through a dataset. Once you have this “wrinkle” ironed out, delete the older component containing those disabled cameras, as the tie points in that Component will inform (pollute) any Align in that scene, a good practice to omit really small components, they’re strong candidates for misalignments, then go back and re-enable those cameras> run align again> RC now creates new tie points for those cameras and uses the strength of a clean(er) Component to bring these cameras back into play.
This kind of thing simply takes lots of practice to get a feel for, stick with it.
Benjy