No, by "the rest of the paragraph" I mean "I am not ... your sources are complete garbage". (I did comment on the carrying-capacity thing, but that was separate.)
An ignorant perusal (e.g., mine) of the Wikipedia article gives the impression (1) that someone who has a clue about this stuff has claimed that those wars were incredibly deadly (suggesting either that subtracting the censuses isn't as hopeless as it sounds, or else there's other reason to think so besides the censuses) and (2) that the people who did the later census -- the one with the smaller numbers -- did try quite hard to make sure they didn't miss anyone (so in so far as they succeeded, the apparent loss of numbers estimated from the censuses would have to be an underestimate).
On 1, WP cites four sources. Two look very unscholarly, to say the least, but the other two seem superficially credible enough. Are they just plain wrong? On 2, a single source is cited that again looks kinda-credible, but I have no way of checking whether the source says "... so subtracting these censuses can't give much too large a figure for the loss of life" or something more like "... but although the Jin did their best, bless their little cotton socks, it seems likely that they unerestimated by a factor of 4". I was hoping that you might answer my questions looking for more specificity, and help me distinguish between the hypotheses (1) that although you can't expect to get useful results from subtracting those census results there's other good reason to think there was a huge loss of life during the Three Kingdoms period, (2) that actually there are ways to get useful information from comparing the census results, and (3) that actually the only reason to think there was huge loss of life is the comparison of census results, which is just as hopeless as it sounds like it might be.
(#3 is very plausible prima facie, but that Wikipedia article seems like some evidence for #1-or-#2. I am well aware that Wikipedia articles can be wrong, hence my request for more specifics that might clear the matter up. There's some interesting but inconclusive discussion on the article's talk page.)
Every discussion of death toll I have ever seen is a simple subtraction of the two censuses.
The two censuses were more than a century apart, in 157 and 280, so they cannot tell you anything about killing. Going in the opposite direction, if the killing is spread out over time, it could easily kill twice population with no net effect.
I think 10% killed counts as "incredibly deadly."
Yes, there is a line in wikipedia claiming that the Jin census was high quality, but other people claim that it was low quality. In fact, what that line really tells you is the existence of people claiming that it was low quality.
It might seem odd to call politics undervalued. But if we want to think about how to improve matters for millions of people, it's clear that effective governance and institutions which promote human flourishing are extremely valuable. One example is the Jin dynasty: in between the end of the Han dynasty and the rise of the Jin, the Chinese population dropped by 4/5ths. The wars immortalized in Romance of the Three Kingdoms ere the deadliest before WWII. The existence of any halfway competent central power would have spared millions of people.
Closer to home housing policies in the Bay Area force the poorest in society to spend vast sums to keep a roof over their head. Poor educational opportunities reduce the ultimate earning potential of millions. In some cases, such as pollution controls, the costs of poor policies are measured in lives.
Many people enter politics, and competition is tough. But many opportunities exist that are frequently underexploited, such as school boards, transportation boards, and down-ticket positions with considerable power. Often these positions are dominated by people without an understanding of the issue, and subject to heavy lobbying from those who are directly involved. Occasionally important policies are made in relative obscurity: who here knew about the raisin price controls before it came before the Supreme Court?
There are many ways in which politics might be affected. One is through political theorizing. Robert Bork's The Antitrust Paradox influence a generation of judges and policymakers, leading to a radically different antitrust policy in the US, which benefited consumers. Milton Friedman was the handmaiden of a major shift in economic policy in the US. On the other end Sayyid Qutb has been blamed for the rise of Islamism, which has made life a great deal less pleasant for many people around the world. However, the returns on political theory are extremely uncertain: many works of political philosophy go ignored.
Another is through political organizing. Here the model is the Social Democratic parties of yore, which had profound effects on the social institutions of the countries in which they operated. To the extent the poorest in Europe are better off because of these parties, this work directly improved people's lives. Unfortunately, there is good reason to suspect this is much less doable today.
A third is through competent leadership. Many municipalities are governed poorly for a variety of factors. Publicly minded citizens with slightly better than average interest should be able to copy good policies into poorly governed cities. Many bad policies are the result of rent-seeking, which can easily be resisted, at the cost of having resources for reelection.
Politics inherently involves leverage. Decisions about policies affect all those in a jurisdiction where the policy applies, and can have knock-on effects, such as financial regulation in a major finance center, or the impact of California emission standards on automobiles. Furthermore, good institutions can last for centuries.
At its most extreme, decisive political action has changed the fates of millions. At its least extreme an effective mayor can ensure that children are educated, potholes repaired, and new housing built. In between politics may offer the best leverage of any opportunity for altruism.