One way that works for one pixel and could work in different ways for multiple pixels...
Summary:
In short create a single point in an editing session, use it as a mask feature, extract the pixel from the source, reclassify it, append it back to it's source (using source as the target), then reclassify it again to normalize it with the rest of the source raster.
General Steps:
1. Set the environment's geoprocessing x, y, z tolerances - this important
2. Create a point feature class, import the coordinate system of the raster, start an editing session, zoom to the pixel and add a point on the pixel and save
3. Use Spatial Analyst Tools\Extraction\Extract by Mask, set the input raster and use the point as the feature mask data and run
You should now have a single pixel in the extracted output raster
4. Use Spatial Analyst Tools\Reclass\Reclassify and assign the pixel the value you want
Here's the tricky part:
5. Use Data Management Tools\General\Append and append the pixel raster to it's source
At this point it may not look like it worked because you can not access the new value in the ArcMap symbology tab, however...
6. Run Spatial Analyst Tools\Reclass\Reclassify then "Add Entry" and type the value in as old and new
7. After the reclassification you should have a raster with the pixel updated.
I was using ArcMap 10.5 with LiDAR derived rasters and used a similar technique to clean up a bunch of pixels in an area using a polygon mask...