Maps Made Easy and all other photogrammetric processing software are generally doing the same things to create a 3D model out of a set of 2D images. By using images of an object taken from different perspectives the shape of that object can be discerned by detecting common features and comparing the angles from which the images were taken. Generally, the more data that is available the better the reconstruction will turn out. If one side of the object has 10 images of it and the other only has 2, one side will turn out better than the other.
Part of a map or model being distorted almost always comes down to a lack of overlap or missing perspectives. Please read here for more information on missing features: Distortion and Missing Features
If the object that didn't turn out well was the roof of a building, please refer the following link since that is a separate topic:
For creating orthophoto maps, it is not necessary to be able to render overhangs and detail on truly vertical areas (that's the ortho- part, it means perpendicular). This is referred to as 2.5D. It is almost 3D, but not quite. Capturing 2.D requires 5-7 views from different perspectives overhead unless the scene is really complex (like with trees) which will require more.
For creating an accurate Full 3D model that includes rendered vertical and overhung surfaces, we generally recommend getting AT LEAST 8 looks at any given face that is to be rendered. This means that for overhanging areas, views will need to be attained from below looking up at the area. Collecting data for Full 3D reconstruction is a very different process than taking mapping data and it is very hard to automate safely.
To read more about the recommended number of views various targets require, read about overlap reports here: Overlap Reports
Limitations in the Technology
We would love to tell you that everything will always be able to turn out perfectly if you take enough images. But as with every technology, there are limits to what can be done using photogrammetry and a camera. Some scenes are simply too complex to turn out well.
As is described in the Data Collection guidelines, if there is any motion with the object that is attempting to be captured, there is little chance it will turn out well.
Due to the nature of our service, you can't clean up the edges after the fact. The only way to do it is to collect better data and reprocess.