Wasn't going to comment on this story, but have been nudged by a friend. So, some thoughts on Afghan data, from someone who's spent the past 7-8 years measuring attitudes and control in Afghanistan (thread) nyti.ms/2NY0J2C
(1) The mismatch between official and external estimates of control is not surprising. It's doubtful that ISAF/RS has had fidelity to estimate district level control since 2011-12. No serious researcher would place stock in ISAF/RS estimates (and hasn't for years)
(2) The district probably isn't the right place to measure control. It's convenient, yes, but few Afghans identify with their district, and it doesn't mean much on the ground. These codings also mask lots of heterogeneity *within* the districts at the village level
(3) Village level estimates of support and control are really what are needed (like the Hamlet Evaluation System). But neither RS nor outside analysts have figured out how to scale up from village level to district and larger (or to build campaign plans for every village)
(4) That said, a quick look at the district map of control reveals that the Taliban now control or heavily influence the entire ring road (Highway 1) thru Afghanistan + entry points to main cities. They don't need to take the cities; they could besiege them (shades of 1994-96) 
(5) I'm a broken record on this, but I think the most telling data are ANDSF retention rates/force size. (BTW, the Taliban casualty data are also highly suspect). The fact is, ANDSF has been shrinking for years. Taliban, though tired, are enjoying ever-favorable battlefield odds
(6) Taken together, you can see why the Taliban are going to continue to press home their battlefield advantages (b/c they're winning) while periodically entering into negotiations (or "pre-talks"). So, more "war, war" even as we get the occasional "jaw, jaw"
