To make this as simple as possible, imagine GT Sport has two cars, CarID 0001 and CarID 0002. The WOM has a list of numbers from 0001 to 0002 and moves from one to the next every time you meet the requirements.
You can easily see how that would, over time, give you exactly what you expect: half the time you get 0001 and the other half you get 0002.
Now imagine that PD planned to expand the game with one more car, so it gave the WOM a list of numbers from 0001 to 0003. Initially, 0003 isn't a valid CarID, so instead of picking 0003 it rolls over to 0001.
That would result in 66% 0001 and 33% 0002. To you that would look like the WOM is programmed to give you more 0001s than 0002s, but that's just a consequence not an intent. Then along comes an update, the car with CarID 0003 is added, and now the WOM gives you each car a third of the time; it hasn't changed at all, but the cars it produces have.
You can see how that gets dramatically larger with a larger list of CarIDs. Sticking with just the two cars but a list from 0001 to 0005 would give you the car with CarID 0001 80% of the time. Add a third car with the CarID of 0004 (not 0003) and you'd get 0001 40% of the time, 0004 40% of the time, and 0002 20% of the time.
That gives the impression that certain cars are favoured - and which ones are favoured changes with content updates - but it's just not the case. It's an emergent behaviour, not something intended.
I don't know the range of GT Sport's list, but I can say that the CarID range is at least 0201 to 3364 (because that's all the numbers I found before I stopped caring) and probably much larger. That's at least 3163 CarID numbers for at most 323 cars (the WOM cannot give you 13 of the cars: the Pace Cars and Gr3/GrB Road Cars), which means that almost one in ten CarIDs on the Daily Marathon list are invalid and will be skipped.
This is almost certainly what generates the wildly disproportionate appearance, or lack of appearance, for certain cars. It looks like the Gr.3 and Gr.4 cars are clustered around certain regions of CarID numbers, so you are far more likely to see either the first or last (depending on how WOM finds the nearest CarID, looking up or down) cars in those groupings than you are to see others in the middle. There may well be other patterns in how the CarIDs are grouped, but I haven't looked any further into it once I realised how largely pointless it would be.