... how would you most effectively distribute them?

Make a list of 100 interesting authors and check their blogs manually on a regular basis. That way you can make 40 of your 50 "impactful votes".
For the 10 remaining "impactful votes" try to find good posts from authors you have never read before to give new users a chance and some motivation to keep writing.

And the modest bump to each one would not provide much exposure (e.g., on trending), which means it's unlikely for others to find it and place more votes on top (and increase your curation return).

I don't understand why investors always talk about their ROI (here in this special case "curation return")? Their main aim should be to increase the value of STEEM. Voting for 'quality content' (and flagging bid bot supported posts) would be a contribution to a higher value ... (with or without much curation reward).

The value of a (social) network is measured among others by the number of its users.
So lets make sure that as many as possible users are having a pleasant user experience (for example also because they might get some impactful manual upvotes from time to time) and thus stay here. Lets do that as investors to save our investment.
If you have one million STEEM it's not most important to get even more STEEM, it's important to increase the value of these STEEM you already own.
For example I have much more STEEM than a year ago, but my account value has decreased significantly.

Did I need 'ROI' when I bought BTC some years ago? Or am I happy about the high value of my BTC nowadays? :)

Nothing against ROI, but I think we shouldn't be that focused on it ...