The Spatial Analyst > Math > Logical > InList tool appears to have a bug in 10.5 that was not present in 10.3. The result of the InList seems to ignore the NoData areas of the input raster. Here is the input layer (soils) that shows NoData values outside of the red boundary.
Here is the result of the InList tool from ArcGIS 10.3.1 - notice the NoData values outside the boundary.
Here is the result in 10.5.1 (10.5 produced the same result). Notice the result seems to extend the pixel values at the edge of the boundary to the edge of the raster extent.
I know that the easy solution is to clip the layer, or use the reclassify tool instead of the InList, but I am wondering if there is something that I am doing wrong. I've tried making a new copy of the input, changing the projections, etc.