The options are purely theoretical of course. We could use the inflation rate to calculate the theoretical distribution curve over time using each of the options, where option 3 could be split in a lot of different subcurves. But what would that show?

I don't have the exact numbers at hand, but roundabout they should be fine:
inflation rate: 10% (for the sake of rounding, it's actually lower)
a) dao 10%
b) witness 10%
c) inflation on SP 15%
d) curation 32.5%
e) posts 32.5%

Self voters receive 80% (c+d+e) of the total inflation and leave 20% for witnesses and the dao - both basically developers and (self-)marketers.

Burning the posting rewards removes e) and changes the remaining as follows:
inflation rate: 6.75%
a) 14.8%
b) 14.8%
c) 22.2%
d) 48.1%

Now we have 70% of the inflation for the burn voters, and 30% for witnesses and the dao. Not a huge change in percentages and still no distribution outside of those two groups.

For option 3 let's assume curators who vote 50% established, 50% new users (which is pretty optimistic). The percentages are the same as for option 1, so:
20% for witnesses and dao, 63.75% for stakeholders and 16.25% for new users.
16.25% of 10% inflation mean a dilution of 1.625%.
With votes going 100% to non-stakeholders that would switch to 47.5:32.5, bringing the maximum possible dilution to 3.25%

Okay, now to the interesting part :D To calculate the real effect regarding dilution of stake, we need to include the vesting ratio. Hopefully more or less accurate numbers I found quickly state that 210M Steem are powered up, at a total of 360M, which results in a ratio of 58.3%, let's say 60. This means that 40% of stake which the inflation rate is based on gives up on their share on c and d, leaving it for the rest to grab. It factually raises the rewards, giving the ones using c and d 66% more rewards than if all stake was active.

1000 STEEM created actually distribute the following way:

option 1&3:
a) 76
b) 76
c) 189
d) 412
e) 247

option 2:
a) 101
b) 101
c) 252
d) 546

So - while option 1 leads to a self-allocation of 85%, option 2 lowers that to 80%, leaving the other 5% to witnesses and the dao.
Option 3 allocates 60% of the inflation to the voters.

Conclusions:

I) The maximum distribution rate if all votes would go to non-stakeholders is 2.5% of the total stake per year, plus 1.5% via witnesses and dao, resulting in a total maximum dilution of 4% per year. All those values are actually lower, because I think the inflation rate is closer to 8% than 10%.

II) Burning post rewards only has a minimal effect on self-allocation of inflation compared to self-voting. In this purely theoretic scenario at least. In reality there are a lot more factors, for example that an individual using option 1 hurts everyone else, while option 2 is neutral in that regard.