In lieu of actually looking at the code of the blockchain I am going to make the conjecture that the oficial Steem White Paper is an accurate description of it.

From page 26 of the Steem White Paper

75% of the new tokens that are generated go to fund the reward pool, which is split between authors and curators. 15% of the new tokens are awarded to holders of SP. The remaining 10% pays for the witnesses to ​​power​​ the ​​blockchain.

From page 17:

When a post receives a pay out it takes the form of 50% SBD and 50% SP.
Users also have the option to be paid in 100% SP, as well as decline payout on posts.

If 75% of the inflation goes to the reward pool and 75% of the reward pool goes to author rewards it follows that the liquid portion of the post payouts comes from that same reward pool.

Now, over time that distribution may not be equal to the intended percentages due to the fluctuations of both the price of the Steem token and the 10% ceiling of debt ratio of Steem/SBD.

I understand that the above reasoning is not the same as showing empirical evidence. We could take two posts as samples. One with an SBD print rate of 100% and one with 0% to see if those percentages hold true. The tricky part is knowing what was the median price feed at the time of payout. But it's late and I need to go to sleep.

Too bad that https://steemdb.com/ is down (maybe it's being decomisioned?). On the front page it had stats showing the actual distribution of the tokens for the last 30 days.