The graph cites a broken link to the US Census Bureau. A few minutes of digging led me to this table. As you can see, the only number for 400 BCE is from Biraben, whose numbers are higher than the McEvedy and Jones numbers the table depended on before. Biraben numbers often are the upper limit, but according to the rules of the table if only one number is given, it is automatically made the lower limit. That would throw off a line graph of the lower limits, which is possibly what the graph you linked used (but the scale is useless).

So I'd say a noncritical Wikipedia editor just took what they had and ran with it.