Thom stated in a catchy metaphor kinda what I was thinking after looking at this yesterday; so I agree. I think it's a way-cool idea, but it just doesn't seem quite ready for prime time.
Reading the methodology page closely, it still isn't clear precisely what is being displayed. Ancient samples are mentioned, and the bulk of the data is being pulled from YFull (which is understandable; the bioinformaticist who's presenting the data had a falling-out with FTDNA some time back).
It seems the heatmap is trying to display contemporary density, not anthropological timeframe origins. The author's main app, BTW, "Mygrations," is quite good (again, as far the available data go) at approximating an animated yDNA anthropological timeline: https://phylogeographer.com/snp-lookup/.
But for the heatmaps, ancient samples really don't matter much: there are so few significantly sequenced ancient remains (most have only limited SNP and/or STR results because they were done circa pre-2015) that they really wouldn't blip the radar on a contemporary population heatmap.
Having been a customer of YFull's for a couple of years, I honestly don't know where reasonably accurate geolocation information could be coming from in their database...unless it is not from contemporary user-submitted data at all. For my paid account, YFull has my mailing address and my subjective estimate of the country my EKA was born in. So for me YFull has USA and GBR on file: where I lived when I purchased the service, and where I think my EKA hails from.
I could have guessed Fiji for my EKA because for none of the SNPs/haplogroups I tried was there ever any results at all on the heatmap in North or South America. Nada. And for many, there are small hotspots out in the middle of the Atlantic Ocean.
More yDNA test takers are in North America than anywhere else, and FTDNA and its database are the industry's 800-pound gorilla for yDNA. There are very large numbers of R1b and, downstream, M269 men in North America. That absolutely is not displayed in the heatmaps. In fact, for M269 and several steps deeper, down to DF27, North America remains zeroed but a perennial hot-spot shows up in the Philippines. Southeast Asia. Almost exactly the opposite side of the world from where actual contemporary concentrations of M269 or DF27 would be.
A separate issue with YFull is they have long maintained some instances of analyses and haplotree branch taxonomy that are in distinct disagreement with FTDNA. As an example--while both SNPs were first discovered and named by FTDNA--YFull and FTDNA have me placed on entirely different branches of R-BY3332. YFull thinks I (and by extrapolation the 36 test-takers in my Williams yDNA subproject) am BY3332 > BY19276 (which is an analog of L881). FTDNA has us as BY3332 > BY22166 > BY22194 > BY35076 > BY35083. So in other words, that haplotree branching occurred sometime around 4,000 years ago and FTDNA and YFull thinks our ancestors took two different branches of the tree at that time. Depending on your individual results, there can be a huge disconnect between the data reported by FTDNA, who does the vast majority of full-sequence DNA testing right now, and YFull, who does no testing at all.
As a final comment, for those who try the utility, keep in mind that it seems to be case-sensitive for the input. Too, valid input is inconsistent. For example, if you want to look at haplogroup G or E, just type "G" or "E". Doesn't worth with others like I, J, K, R, etc. "R1b" works, "R1b1" doesn't. L151 is barely there; its parent branch, P310, is nonexistent, while its child branch, P312, shows high densities throughout Europe...and, of course, the Philippines. So if at first your searches don't succeed, try reformatting the input query, or moving up or down the particular haplotree branch. And take the results with a grain of salt.