While the rewards on a post are split 50/50 per the code, there are various mechanisms, such as DV's, and curators targeting higher valued posts, the reduced author rewards for comments introduced in HF27, and now the change in HF28 that more rapidly bleeds VP enabling whales to more efficiently maximize curation rewards while discouraging engagement by bleeding minnows dry of VP when upvoting comments, as has recently happened to me and left me clawing and scrabbling to rebuild my VP for more than a week now. On top of these overt mechanisms, circle jerks, self votes, botnets, and other rewards pool rapine continue to operate more covertly to focus the rewards pool on whale accounts and increasingly exclude minnows and lesser staked accounts.
Cumulatively these mechanisms contribute to whales extracting >90% of rewards from the pool, curation rewards being ~60% of outflow from the rewards pool, and author rewards ~10%. 10% of the pool goes to the DHF where the largest outflows of the DHF are operations whales almost exclusively benefit from, such as the HBD stabilization program, Valueplan, and etc. For more explicit understanding of how whales almost exclusively extract rewards for HBD stabilization, please look at HBD-funder comments, which are employed to apply author rewards to the HBD stabilization fund, but actually mostly are paid out to the voters on those comments, which are published 20 at a time or so, as curation rewards. 10% of the pool goes to witnesses, and the best rewarded consensus witnesses are mostly whales, and have been consensus witnesses almost continuously since 2016.
Edit: there are details of curation/author split that require deep investigation but also contribute to curation rewards dominating rewards pool outflows, such as that upvotes on creation after 1/2 hour (IIRC) provide authors a percentage of the curation rewards in addition to author rewards, and I have little familiarity with more esoteric mechanisms that matter a great deal to substantial stakes seeking to maximize ROI.