Frankly, with as many images as you are combining, you won't see much difference with most of the different methods.
But what method one should use depends on a lot of factors.
If you have done things right (and you have), you have lots of images to use. I think I see 24 images in one set, and 25 in another set. (Why is it that you don't just consider it 49 total images?)
With that many images, I would be using a sigma combined in IP. (Assuming you have sufficient exposure, and with 8 minutes at ISO you should have.) Sigma combined runs a statistical analysis of all the values for that pixel in the array, and then rejects all the values that are not within a given standard deviation of the mean, while combining the rest. That means that all values are considered, and the center is weighted.
Median combine instead uses only the median value. One value out of the twenty four possible values. Now, with twenty four exposures, this has got to be pretty close to a true middle of the pack. However, let's say you only had three exposures, and two of them had a hot pixel, or a satellite going through, or a cloud passing by-- your "median" would not be truly representative.
A simple average combine would overcome some of these problems, but let's say you had three exposures, and one of them was bad. That one bad value would have one third the data you used. It would be averaged out, but still have lots of effect on the end result. Again, if you have twenty four exposures, it is not that big a deal, since the bad pixel value would represent only one twenty fourth of the data.
In short, You should always try to have at least eight exposures, and then use a sigma combined.
If you have fewer, you can try one of the other methods. If it is particularly dim, you can either straight add (and not divide by the number of exposures) or try the funny add (adaptive add???--I cannot remember the name) where you change the divisor after you add.
However, at 24, you reach the point of diminishing returns.
Your mileage may vary.
Pretty picture, though.
Alex