For example there are about 10,000 papers submitted per month to all of arxiv[1] - which is a huge number. That totals 132k per year or so. For there to be "dozens of millions" of journal articles, that would mean arxiv has just 0.83% of them. Given the very low bar to publishing on arXiv I would think if you compared the number of arxiv articles as a percentage of all pubished journal articles, it would be a lot more than 0.83%! So the "dozens of millions" of journal articles is extremely astounding to me.
A Nature study from 2014 (http://blogs.nature.com/news/2014/05/global-scientific-outpu...) pegs the number of papers published between 1980 and 2012, with at least one citation, at 38 million. So, 69 million publications including uncited ones and ones from outside this time range actually sounds like an underestimate - that is, the 69 million paper archive is probably missing a fair number of articles.