The basic point of those diagrams is correct: with larger rectangular dies on a round wafer, you can’t fit as many complete sensors, and edge areas often can’t hold another full-size die in the required shape and position.

Using leftover spaces for smaller sensors is not inherently impossible. In principle, manufacturers can mix die sizes on one wafer. But in practice it adds manufacturing complexity and cost. Different layouts can require different probing/test setups and may interrupt an optimized production flow built around one design. Edge regions are also less desirable because mechanical processing near the perimeter can introduce more stress and defects, so that area often has a higher failure risk.

There’s also a yield issue: larger sensors are more likely to be ruined by a single defect, so losing one large die wastes more wafer area and value than losing one small die.

So the illustration is simplified, but not fundamentally wrong. It shows the geometric and yield penalty of large sensors. Filling gaps with smaller sensors may be possible, but it’s not always economical or practical compared with running wafers dedicated to one sensor design.