Jpg, png, and other kind of images are already compressed so there is tiny or no advantage at all in using a binary loseless compressor.
The best setting is probably
Compression Method: PPMD (instead of LZMA)
Compression Level: Ultra
However the big deal, the real trick that makes a huge difference (when you can apply it) is sorting the images renaming their names of gathering them into directories so to have a sequence with the lesser graphic difference possibile amongst them. The compression is the maximum possibile when it's like having a video clip with little difference among frames.
Just to highlight the result: I've compressed 900MB into 260MB! The images were a series of backgrounds for websites sorted by pattern and by color.
If you have a series of photos from your camera, they are quite surely already alfabetically sorted, but this doesn't automatically means they are well sorted for this purpose.
Let's say you have a series of photos 1 to 9 with subjects snapped in this order: garden1, sky1, water1, garden2, water2, sky2, garden3, sky3, water3. You would surely have a better compression renaming the file name so that the subject order was garden1, garden2, garden3, sky1, sky2, sky3, water1, water2, water3. That's because the sky* are 3 photos with minor difference among them compared with the garden* full of green. And the water near the sky is better than having sky* gardern* water*
Again: the most pixel are similar in the sequence, the better.