Because shipments are not sales, and the U.S. isn't the whole world. It's not hard to understand. Sales can be up in a given quarter even if shipments are down. Let's take a look at PS2 shipments vs. sales during the system's prime sales years (2000-2005), since I can easily find quarterly shipment data for it. Specifically, let's look at America. Sony's shipment data covers North America, while NPD data covers the U.S. as a whole, but generally speaking the U.S. represents on average some 85-90% of the U.S. market, and, aside from some screwy exceptions (i.e., supply shortages and other disruptions) shipments tend to outpace sales, for obvious reasons.
Sales and shipments seem to rise and fall in tandem, right? Maybe at first glance, but 2004 clearly shows that sales in a given time period can outpace sales. But just to be sure, let's break this down by quarter.
Running into a bit of a problem, aren't we? While we see the clear and familiar pattern of sales remaining low during non-holiday quarters and then experiencing a big jump in Q4 (which averages around 50% or so of the year's sales for most systems), shipments don't really exhibit a clear pattern. In 2001 to 2003, Sony ramped up shipments throughout the year, peaking in Q3, while Q4 shipments saw a drop from Q3, almost certainly because they already had plenty of stock in the supply channels ahead of the holidays (we've seen non-holiday global shipments outpace holiday shipments for the PS4 as well), but after 2003 we don't see that pattern (the transition to the PS2 Slim, which was plagued by early shortages, was a likely culprit). We see several incidences of sales in the U.S. being greater than shipments to the entire region, both in holiday and non-holiday quarters. We see at least two quarters (Q1 2002 and Q4 2005) where sales were up YoY while shipments were down.
And this is in just one region. If I had the time (it's taken me over an hour to make those charts and write up this post), I could show the same for Japan, and for systems besides the PS2. I could also show that global shipments in a given quarter can vary wildly in what share of those shipments each region gets. So, once we add in the complications of global shipments vs. global sales, we clearly go way past any reasonable ideas that shipments are in any way congruent with sales.
TL;DR, it is not at all unusual that we could see global shipments for the PS4 decline in Q2 of this year despite sales going up in the U.S., staying statistically flat in Japan, and probably not dropping by any appreciable amount in Europe, either (if VGC is any indication, they're actually up quite a bit in Europe as well). It always bears repeating: Shipments are not sales!