I think this is a very interesting question (and also think it's funny how no one has really addressed it). I don't know the answer decisively, but my hunch is that the ad format shown is based more on what ad buyers are willing to buy for a given video.
YouTube is very vague on this, but the YouTube help page on default ad settings does say this:
When you monetize a claimed video, YouTube displays advertisements with the video (when users view it on a monetizable platform). You specify which types of advertisements YouTube displays by selecting the permissible ad formats. YouTube automatically determines which ad format displays for any given viewing based on a variety of contextual factors.
So, the link to the "variety of contextual factors" is also super vague, but the relevant quote from this page is:
The Adsense ads displayed on your video are determined automatically by our system based on a number of contextual factors relating to your video. These factors include but are not limited to your video metadata and whether the content is advertiser-friendly.
I feel like this supports my hunch -- YouTube has an interest in posting the most lucrative ads on every video it can (which would typically be the video ads, as you yourself note)...but it needs to be able to convince advertisers that a particular video is worth putting a particular ad on. So I would guess that to the extent videos don't have the video ads, it's because YouTube either can't sell those ads for the particular videos, or it may just be a matter of YouTube trying to maximize it's revenues (e.g., it can't get video ads all the time, so it gets video ads some times, and then fills in the gap with other types of ads.)
So, let's say that YouTube can only place a video ad 50% of the time, and then the other 50% it puts overlay ads. If you select video ads only, I don't think that means that it will put twice as many video ads. Rather, I think that means you'll have the same amount of video ads, but no overlay ads (and thus miss out on the overlay revenue, even if it's less than video ad revenue per ad.)
But as I said before, I'm not an expert here. This is my hypothesis.