An extraneous character was my thought also, with the usual suspect being a newline, the data being read from a file (but in that case, the regex must be modified with  /s so that "dot matches all"), or perhaps some kind of "end-of-text" character. Did you do a hex dump of the file yet?

>perl -wMstrict -le
"my $s = qq{pl\n};
 ;;
 print 'match sans /s' if $s =~ /p.{2,6}/;
 print 'match with /s' if $s =~ /p.{2,6}/s;
"
match with /s
[download]

perl -e 'print "pl\r\n"' > /tmp/j.txt

perl -e 'open(I,"/tmp/j.txt");$/=undef;$_=<I>;
  print ". matches CR on $^O\n" if(/p.{2,6})'
[download]

I can reproduce what you describe. Firstly, two text files which only contain the string 'pl'.

$ cat junk
pl
[download]

$ cat junk2
pl
[download]

One matches your regex, the other doesn't:

$ perl -nE 'say +(/p.{2,6}/) ? "match" : "no match"' junk
match
[download]

$ perl -nE 'say +(/p.{2,6}/) ? "match" : "no match"' junk2
no match
[download]

A closer look inside the files explains why.

$ cat -vet junk
pl^M$
[download]

$ cat -vet junk2
pl$
[download]

[In case you're unfamiliar with the command, cat -vet, displays carriage returns as "^M" and newlines as "$".]