I would argue that cloud cover has more of an impact the less sea ice we have, but this becomes especially pronounced as area decreases relative to extent (i.e., as compactness goes down).
Consider; in 2012, things were bad, but the main pack was still cohesive. Whether or not this area was covered by clouds, it appeared entirely (or almost entirely) white to the satellites, with the uniformity in concentration aiding "steadiness" in observations. Cloud cover over open ocean is easier to reduce than cloud cover over a background that is the same color.
This year, with structure of the pack completely shattered, you have -- for the very first time -- only a small area of ice that is structurally intact. While this small triangle of area N of the CAA/Greenland falls into the same category as 2012's ice, the remainder most certainly does not.
With ice that is very dispersed, as in this year, cloud cover -- especially of the low sort -- masks much of the open water in between the ice, and the satellites have a much harder time distinguishing what is real vs. what isn't, at least when clouds are overhead. Additionally, I think open water *encourages* cloud cover this far N, especially as it gives up its latent heat from the summertime, further reducing accuracy as clouds have been nearly continuous over much of the Arctic this summer.
Again, the key is the overall structural integrity of the ice.