The agreement with the theoretical curve is indeed remarkable, and perhaps too good. Do you remember school lab experiments doing things like measuring Ohm’s Law. There always was a great temptation (for others!) to adjust their results to better fit with the (“known correct”) predicted results. If that happens with no financial considerations, imagine how much pressure there is in these Big Science experiments where the future funding is critical to everyone’s jobs.
Does anyone remember Diesel emission testing? Again big money involved, and software tweaked to get better answers.
I don’t know, I just have a concern. There was some guy (whose name I have forgotten) who wrote about the results with great passion, although he did have an obvious agenda. Nevertheless he stated that the COBE results did include (to the best of my recollection) something like 1000 correction terms. If you know the “correct answer” it is pretty easy to find errors which move the result in the desired direction. This can easily be an unconscious bias, so I am not accusing anyone of malpractice. It is just a concern of mine.
Then there is the problem with a null-sensing system at low cryogenic temperatures. Whilst I have never myself worked at cryogenic temperatures, I did design, build, and test room temperature mmwave radiometers for a living for 12 years. The equations suggest that they get very insensitive at low temperatures. This is a non-obvious result, and it is not clear if this was known to the investigators.