The main argument is that at scale NXT POS would anyway be transformed to Bitshares' DPOS because of economies of scale. The argument is: If thousand of tx per second have to be processed by NXT forgers they would anyway delegate their forging power to some specialized entity (data center or someone with a powerful server) because of the bandwidth requirements etc. 

The other argument is that you can reach much more tx speed if the number of block producing nodes are fixed (DPOS targets a 1 sec block time, starting with 5 sec in Bitshares 2.0).

Slightly related: https://nxtforum.org/general-discussion/nxt-pos-vs-bitshares-dpos/?PHPSESSID=7doicu7e2bnortfuvfjdhakp02