We report a new free-floating planet (FFP) candidate, KMT-2017-BLG-2820, with Einstein radius $theta_esimeq 6,muas$, lens-source relative proper motion $mu_rel simeq 8,masyr$, and Einstein timescale $t_e=6.5,$hr. It is the third FFP candidate found in an ongoing study of giant-source finite-source point-lens (FSPL) events in the KMTNet data base, and the sixth FSPL FFP candidate overall. We find no significant evidence for a host. Based on their timescale distributions and detection rates, we argue that five of these six FSPL FFP candidates are drawn from the same population as the six point-source point-lens (PSPL) FFP candidates found by citet{mroz17} in the OGLE-IV data base. The $theta_e$ distribution of the FSPL FFPs implies that they are either sub-Jovian planets in the bulge or super-Earths in the disk. However, the apparent Einstein Desert ($10latheta_e/muasla 30$) would argue for the latter. Whether each of the 12 (6 FSPL and 6 PSPL) FFP candidates is truly an FFP, or simply a very wide-separation planet, can be determined at first adaptive optics (AO) light on 30m telescopes, and earlier for some. If the latter, a second epoch of AO observations could measure the projected planet-host separation with a precision ${cal O}(10,au)$. At the present time, the balance of evidence favors the unbound-planet hypothesis.