About half of those under 30 (53%) report having ever used a dating site or app
Yes, but that's a bigger denominator, and includes single people, and even those who have never been on a date. The headline question is what percent of couples met through different methods, not what percent of individuals, including those who are not currently in a couple.
So it doesn't make sense that more people would have met their current partner through a dating app than have ever used one.
It could be that a higher percent of couples met online than the percent of people who have ever used online dating. If you have a data set where online dating is literally the only way to meet people, but only half of the people are trying that method, you'd have the situation where 100% of couples met online but only 50% of people have ever tried online dating (this hypothetical is purely to demonstrate the math, not claiming that this is in any way a reflection or the actual data).
It's entirely possible (and I'd argue is likely) that the 53% who have used dating services are more likely to be in couples than the 47% who haven't. And so that larger subset of the 47% would therefore be excluded in the "percent of couples" data.
mostly I'm just bothered by the apparent lack of any way to confirm the authenticity of the graph and its relationship to the source material
The 2019 paper I've linked is authored by the maintainers of the linked data set, and contains a very similar graph with an earlier cutoff (2017 data). I'm sure those authors know their data set. It's just most of their papers using this data is paywalled, and the data is mainly used for other types of analyses.
If I have time I might be able to download the data set from a computer and just map it either naively or by applying the correct weights.
You're still not slicing thin enough.
If 60% of the couples who got together in 2022 met on dating apps, and people who got together in 2022 constitute 5% of all couples, that's still possible (and probable), then those couples will still only be 3% of the total. Pretty easy to add up to 11% that way when you start including all the 10-year-old relationships, the 20-year-old relationships, etc.
If it were flat at 60% for all years then no, it wouldn't add up.
But if you look at the area under the curve, it's still pretty small comparatively speaking because it's such a recent phenomenon. (And not every year would actually count equally for the whole data set, but it's displayed in this chart as every year adding up to 100% for its own year.)