I think of it more like you described, each player has a distribution of potential outcomes.
(using "lines" as a proxy for value, you could just use WAR or something like that instead).
The best players peak probably at say a second line level, with a decent probability of a 1st line outcome but also a 3rd/4th line outcome and even below replacement.
As you move down the draft, the distribution shifts down, but varies for each players, some may have a peak at 3rd or 4th line with a small tail, some may have dual peaks, 2nd/4th line with a substantial below replacement tail, and so on.
Past #100 or so, most players have a very thin tail toward the best outcomes, and a peak in the 3rd/4th line area that shifts down as you get deeper into the draft. A few have dual peaks, maybe a small 2nd line outcome paired with a much bigger 4th line/sub replacement value peak.
Now once you get to below replacement level there's no "negative" value, you hit or you don't.
When the chance to get a top two line outcome becomes close to zero, then it makes sense to focus on expected value over "best outcomes."
That is, you may rather have a top 4th line guy with significant probability than a so-so 3rd line guy with a much lower probability, because in both cases they're building depth and saving money, but if you're a good team, you're looking to replace them down the road. And of course you might start drafting with specific outcomes in mind, are they suitable for the PK, an energy line, etc. that would provide more value at the bottom of the roster.