Beleive me, when someone starts quoting percentages start asking questions, as one of the replies earlier suggested, 3.5 is a small number untill you realise it represents 238 million people.
By comparison 97% of 107 is a very small number indeed.
The 3.5 is indeed a very big number, I think most of us on here are very well aware of that.
The 95/97/98% refers to all the samples they've ever tested over at least the last 6 months while they refined their assays (we have no idea how many samples that percentage is drawn from), not to the published study.
And yes, when you see a percentage, you need to know what that's a percentage of and how reliable it is - I tend to think that's obvious enough not to need stating, but perhaps some people really do need to be constantly reminded of that, tedious though it is.
But in my experience, whenever I hear the cliche about 'lies, damned lies and statistics', it's almost never used by somebody who's just investigated the statistics and is about to tell you why they're misleading. It's nearly always used by somebody who doesn't like the statistics and wants to sow doubt about them. The pendulum of scepticism over statistics has swung way too far the other way in my opinion, such that we don't seem to believe any of the numbers any more. Surely there are 'lies, damned lies, misleading statistics, and revealing statistics'? We have to work a bit to figure out which is which...