This strikes me as a straightforward industrial organization anomaly question. Like, why do hot dogs and buns come in different counts to a package or why are the finest apples shipped out of apple producing regions?
Gentzkow and Shapiro (2010) that this paradox is only an apparent contradiction. They find slant is mostly a product of readership bias.
By contrast, we find much less evidence for a role of newspaper owners
in determining slant. While slant is somewhat correlated across
co-owned papers, this effect is driven by the geographic clustering of
ownership groups. After controlling for the geographic location of
newspapers, we find no evidence that the variation in slant has an
owner-specific component. We also find no evidence that pressure from
incumbent politicians or the tastes of reporters are important drivers
But jmbejara already mentions knowing about Gentzkow so it isn't obvious what he's looking for beyond this insight.
Timothy Taylor, Managing Editor of the JEP and blogger at CONVERSABLE ECONOMIST also has an article on this subject that may be of interest,
The Economics of Media Bias.