For an image of 250 cells/inch and a 1:24000 scale the cell size in
meters is
Cell = 24000/250*0.0254 = 2.44 meters
Given that cell size the maximum RMS is 1.2, in this case the total RMS
is 1.05 so this RMS is acceptable.
It is clear that georeference a DRG file it is easy due to the accuracy
of the original image. This image is produce from the original digital
files so it is virtually distortion-free. In addition the markers for
the control points are clearly established in the image which reduces
the likelihood of errors due to misplaced control points.
In this case, the image is from a scanned map. Here two mayor drawbacks are present: first, the scanning process by itself produces distortion in the image, depending on the quality and resolution of the scanner, and secondly, the markers of the control points are not as sharp as the markers in the DRG image so the likelihood of errors due to misplaced control points is higher.
This job is the most difficult of the three. In the case of aerial
photos distortion is present in the vertical and horizontal axes due to
the photographic process and elevation differences in the scene. When
you have a aerial photo the best approach is to orthorectify it but for
this process it is necessary a Digital Elevation Model (DEM),
specialized software such as ERDAS and the camera and fly
characteristics.
An alternative approach is to rectify the image using a good target
layer and a high number of control points. Here my experience have
showed me that a polynomial transformation of order greater than 1 is
better, however, the distortion of the image increases with the order
of the transformation so a balance between positional accuracy and
distortion must be made. When you are rectifying a photo your accuracy
increases with the number of control points you identify.