Bent raster after Build Overviews for mosaic dataset using ArcMap?

Question

I created a mosaic dataset from a set of 12 tiff's spanning the united states. Before running Build Overviews, the images all join together nicely. But after, they look like this:

Any idea what is going on? Did i do something wrong? Should I have merged the mosaic to a single image first?

Per the comment by @PolyGeo, here is a detailed step by step:

1. Gather source data from the FAA (The 12 files under High Altitude Enroute (U.S.) GEO-TIFF Files)
   1. Unzip each zip file into a directory
   2. Each of these tif files have an associated html file that shows some metadata, specifically:

Geodetic_Model:
    Horizontal_Datum_Name: North American Datum of 1983
    Ellipsoid_Name: Geodetic Reference System 80
    Semi-major_Axis: 6378137.000000
    Denominator_of_Flattening_Ratio: 298.257222

2. Open ArcMap (10.3.1.4959 License type Advanced), then create a new Blank Map
3. Under catalog, right click on any gdb, choose New -> Mosaic Dataset...
4. On the Create Mosaic Dataset dialog:
   1. Under Coordinate System, choose NAD 1983
   2. Click Environments...
   3. Under Raster Storage set Pyramid resampling technique and Resample to CUBIC
   4. Click OK
5. Once the dataset is created, right click it and Add Rasters...
   1. Under Input Data choose Dataset then browse for the 12 tif files from 1.1 above
   2. Set Environments... the same as 4.2 above
   3. Click OK
6. Then, I created feature polygons for each tif by clicking through the map file at all the corners of the useful data and used that to modify the footprints for each raster (note that I also set Always Clip The Image To Its Footprint)

At this point I have a nicely stitched together mosaic dataset. But when i zoom out past 1:1,250,000 it stopped rendering the image. I analyzed the dataset which provided the warning that I was missing overviews. So i went ahead and right clicked the mosaic dataset, then Optimize -> Build Overviews. The result, is what you see in the attached image...

Answer 1

I'll bet it is related to your coordinate systems. I think you should review your workflow, paying particular attention to the coordinate systems of your data, data frame, and anywhere a tool used gave you the option to choose one.

Your comment confirmed my expectation:

you were right... The source data was using a custom coordinate system that was slightly different from the coordinate system i chose. So i reimported the images with their natural coordinate system, saved it as a favorite, then redid the whole process always setting to the custom coordinate system.