I had this impression as well, but there's something else going on. If you run the numbers, at T=293, there's a negligible fraction above 5 km/s. The atmospheric loss models typically use an exosphere temperature of T=1000, but again, negligible fraction above 10 km/s.
However, there were some contradictions I couldn't immediately understand. This source [1] says "Only about one in a million helium atoms is lost from Earth via Jeans’ escape.". But it also cites a plot from [2] showing the rule-of-thumb for Jeans' escape: 1/6th the escape velocity vs the RMS thermal velocity. In this case, He has an RMS velocity of 2.5 km/s at T=1000, which is greater than 1.9 km/s.
Agreed the article is not very clear. From the context of the surrounding paragraph though, I think it means one in a million helium atoms is lost via Jeans’ escape every year. Which would obviously be quite a significant factor on any geological timescale.
However, there were some contradictions I couldn't immediately understand. This source [1] says "Only about one in a million helium atoms is lost from Earth via Jeans’ escape.". But it also cites a plot from [2] showing the rule-of-thumb for Jeans' escape: 1/6th the escape velocity vs the RMS thermal velocity. In this case, He has an RMS velocity of 2.5 km/s at T=1000, which is greater than 1.9 km/s.
[1] https://geosci.uchicago.edu/~kite/doc/Catling2009.pdf [2] http://geosci.uchicago.edu/~kite/doc/Catling_and_Kasting_ch_...