Those that want to report that omicron is milder focus on overall stats and average symptoms, and gloss over confounding factors.
Those that want to report that it’s just as severe as delta seem to be just picking up the status quo of poor science journalism, and equating “insufficient evidence” with “conclusive there’s no difference”.
This preprint from today out of SA lays out some prospective good news, but still with a lot of uncertainty over how much the lower severity is intrinsic vs mediated by prior infection/vaccination. And there are plenty of limitations, not least of which that their study only includes confirmed omicron infections up to the end of November (they study hospitalizations, and even had to prune their dataset because some patients are still in hospital).
There is just legitimately not enough data yet to answer the questions that most need answering.