DSSR is behaving as designed. Please see the section "Identification of nucleotides" of the
2015 DSSR paper:
A nucleotide is identified if a residue contains at least three base ring atoms and the root-mean-square deviation (rmsd) of the fit falls below a user-definable cutoff. Since base rings are rigid, the rmsd is normally <0.1 Å. To account for experimental error and special non-planar cases, such as 5,6-dihydrouridine (H2U) in yeast tRNAPhe (Figure 2), the default rmsd cutoff is set to 0.28 Å.
The default DSSR cutoff values are based on extensive tests in real-world applications. Any unidentified nucleotide is almost always due to heavy distortions in its base geometry that is 'beyond recognition'. For example, G248 in your attached
6nd42.pdb file has the PyMOL rendered image as attached. Note the N1-C2 distance is 2.2 Å, far larger than ~1.5 Å (the normal covalent C-N bond length).
DSSR Pro has provisions to handle extreme cases like yours.