The pdb file 1pxx uses insertion nomenclature. This is usually used if for a particular family of proteins a consensus residue numbering has bin defined, but later members were found that have longer sequences. Inserted residues have the same residue number, but a letter in column 27 indicates that this is an additional inserted residue and distinguishes multiple inserted residues. (https://www.wwpdb.org/documentation/file-format-content/format33/sect9.html)

If you open file 1pxx.pdb in a text editor and go to rows 1748 and following, you will see that it is not ASN 105 & ILE 105 but ASN 105 & ILE 105A. If you look at the two residues in a molecular viewer such as PyMOL, you will see that they are subsequent residues in the same chain.

Thank you for your answer dear Honegger. Now it's clear to me.

Using insertion nomenclature has the advantage that, in comparing structures of closely related proteins, the corresponding residues have the same residue number across the different structures, independent of unresolved residues at the N-terminus and insertions or deletions within the sequence. A prime example is KABAT or CHOTHIA numbering of antibody variable domains. However, this may be troublesome for automated processing of structures. An additional problem may be found especially in very old PDB files, as some authors used the alternative localization field (column 17) instead of column 27 to indicate insertions. If you are working with a program that has problems dealing with insertion nomenclature, you may need to renumber the residues sequentially (e.g. https://swift.cmbi.umcn.nl/servers/html/renumb.html , http://www.canoz.com/sdh/renumberpdbchain.pl , http://www.bonvinlab.org/pdb-tools/ ). Many molecule viewers also include options to renumber residues.