The casualty information Patton received would have been synthesized from intelligence collected at a lower level.

This intelligence could come from a variety of sources. Captured personal papers and official documents, prisoner interrogation, and literal bodycounting would give battalion and regimental S-2 (intelligence) officers and other intelligence-gathering units like the Field Interrogation Units and the Military Intelligence Specialist Teams an idea of how many casualties the Germans were taking in their area of operations.

Speaking very generally, this information would then be sent up to the G-2 (intelligence) departments of the division. The division G-2 would compile and analyze the intelligence further and pass it along to the corps G-2. The corps G-2 would to the same and pass it along to the G-2 of its parent army (in Patton's case, this was the Third Army).

As for the quality of the estimates? The Americans had the great advantage of being on the offensive for most of the war in the ETO. This meant they captured large numbers of POWs to interrogate, found German papers and records, and overran battlefields before the Germans could move their casualties. That made the job of intelligence officers easier because they had more to work with. I think it's fair to say the estimates of casualties were generally in the right ballpark.

This is not to say there were not sizable failings in American intelligence. The total failure to recognize the German preparations for the Ardennes offensive being the most notable example.

If you'd like to read more on how intelligence worked in the ETO, this post-war U.S. Army report is quite detailed: https://usacac.army.mil/sites/default/files/documents/carl/eto/eto-012.pdf

​

​

​