In irfanview
png + compress = 2mb
png + compress + grayscale = 1mb
next reduction in size comes from
png + compress + 16 color + grayscale + dither = 500kb
the next increment does change the image ever so slightly, but still not resulting in a bad image
png + compress + 8 color + grayscale + dither = 334kb
You can likely go lower, or you could expand the images to 200-400% and take it even lower, in the 3-4 color range and come back with a file that's just about the same file size and let browsers/image viewers do the heavy lifting of scaling it down.
just for future reference, if you have scans of this quality where you can't see individual dots, but you have dots and not smooth greyscale you can typically do
png + compress + 16 color + grayscale + dither + best color
to every single page and it will need no brain power at all, going lower then this requires a per image check and likely isn't worth the time. photoshop may let you select white and black as colors, and if you can select specific colors then you may be able to automate down to 4 colors with little issue. I find any lower then 10 typically changes things in ways that have to be hand done.