okay. it is a lot simplified, but mostly correct. ideally image format for drawn out stuff and other flat animated stuff is svg (vector graphics - ie - infinitely scalable yet crisp), but png is usually used because it is defacto lossless standrad. lossless here roughly translates to - sensor produced a matrix of colors - lossless photo preserves all data. lossy discards some data. For irl stuff, usually lossless is overkill for end user, hence you see jpegs (defacto lossy standrad)

jxl can so both. others can do that as well. jpegs can be lossless, but that is usually not the standard we use. you can store lossy data in pngs, but the loss is not created by png. jxl behaves by default like lossless (like png), but due to newer algorithms, size when lossless is closer to jpeg. if you prepare loss jxl - it can be close to half size of jpegs.

there are other benifits to jxl (extreme future proofing (extremely high bit depth, and pixel size limit, large amount of channels), progressive decoding, etc.), but our reality has to suck because of google.

I locally use jxl to store family photos, but this means i can not send them, because they are using stuff which does not support jxl, so have to convert and share.