But, yes, which reward curve parameter has been finally chosen?

The parameter is 2e12. It is defined here and implemented here.

where did the 20 STEEM reward pivot point come from?

very interesting question.

Lets says R_B20 and R_C20 are the reward balance and recent claims respectively in HF20. And R_B21 and R_C21 the same values in HF21. The question is, given certain amount of claims, what is the difference in the payout between HF20 and HF21? what is the point where the results are the same?

The relation in the payouts between HF21 and HF20 is

The reward balance in HF20 takes 75% from the inflation, but the HF21 takes 65%. On the other hand the relation between the recent claims in both cases is related to the "claims per rshare" that I described in the post. Then after some calculations we get:

to simplify calculations, lets define k as

The payouts between HF20 and HF21 will be equal when this relation is 1. Then

In my analysis I found that C_R is equal to 0.9. Then k is equal to 0.963. And when we apply this number to the formula, we find that rshares = -56s... in other words, never. So, in my calculations all posts will receive less rewards in HF21 no matter the value of the post, and this is due the introduction of the steem proposals, which are funded from the reward pool.

In order to get 20 steem, C_R needs to be around 0.7. But 0.7 is very low, I ran again my code, and found that right now its value is around 0.868. I think that @steemitblog is not taking into account the 10% from the steem proposals.