I can disprove that chart with one simple concept. The sample size is incredibly small.
Drawing a conclusion about a population based on a sample that is biased, or chosen in order to make it appear the population on average is different than it actually is.
www.logicallyfallacious.com
In psychology and neuroscience, the typical sample size is too small. I’ve recently seen several neuroscience papers with n = 3-6 animals. For instance, this article uses n = 3 mice per group in a …
garstats.wordpress.com
The chart documents the casualties of ~1500 soldiers. That's not even the results of a medium-sized battle, more like a large skirmish.
You can't base the trends of an era on the results of a skirmish, or even 5/10/20 battles.
You need long-term data for accurate long-term understanding.