Give it another 10 years, and then re-visit the subject. :)

In contrast, we show in a prior paper and our new paper that ice sheet models are far too sluggish compared with the magnitude and speed of sea level changes in the paleoclimate record. This is not surprising, given the primitive state of ice sheet modeling.

This needs to be taken in context.

The models used have been written over many years, by many people, at the UK Met Office, for running on several generations of supercomputers.

These people, and those providing the data sets that we run, (and getting the results that we return), are professionals, and are always looking at and for research results in the climate field.
So they'll be well aware of what your post refers to. And a lot more that doesn't make it into the popular press.

As for what data is included in different runs of our modelling, that's up to the researchers running each experiment. And if you watch the files being down loaded onto your computer, you can see some of the data ideas (for want of a better word), that are going to be used.

And the models are getting more sophisticated, as computers available to the public get more powerful.
e.g. You'll notice from the Server Status page that some use the MOSES II land scheme, and some also use TRIFFID.
The earlier models used MOSES, the original scheme. These II versions are using the latest, more detailed descriptions.
(MOSES and TRIFFID are descriptor files for various land use. e.g. vegetation, large bodies of water, high snow forming mountains, etc.)

If you look at this old one of mine under Perturbed Parameters, you can see that it took into account Sea ice, and also the file in which that was contained.
Others over the years have labels to do with ice fall.

As for what ANY military group is using, I'd prefer not to comment. I don't want people with fedoras, dark glasses, and turned up collars coming around and asking me what I'm up to.

Hi Les, Iain, Napoleon ... thank you for this information, excellent posts.

Here is some interesting information from 2013 -- the Guardian --

http://www.theguardian.com/environment/earth-insight/2013/dec/09/us-navy-arctic-sea-ice-2016-melt

The project, based out of the US Naval Postgraduate School's Department of Oceanography, uses complex modelling techniques that make its projections more accurate than others.

sorry for the Long copy and paste.

Some CPDN models forecast the future, or rather "a future", but many models also attempt to understand what caused the past as it is known to have happened: in those "hindcasts" the ice extent would presumably be taken from observations - in which case it doesn't matter that the extents were worse at that time than was forecast at some earlier time.

For forecasts, underestimating the ice loss would be expected to produce a "not quite as bad" world. If the people in that world were to address even that not-quite-as-bad world then they would be better prepared for the worse world that is almost inevitable given our catastrophic indolence.