not tested as you didn't post an example, use this only as a starting point:

URL=http://www.ncbi.org/pub/genomes
GENOMELIST=E_coli.fa.gz E_coli_strain2.fa.gz
INSEQFILE=myLocalFastaFileToBlast.fa
mkdir download
mkdir filtered
mkdir blast

for i in ${GENOMELIST}; do
  wget ${URL}/$i -O download/$i;
  gunzip download/$i;
  faFilter -minSize=1000 -maxSize=3000 download/$i filtered/$i;
  formatdb -i filtered/$i -p F;
  blastall -p blastn -i ${INSEQFILE} -o blast/$i.blast -e 0.000001;
done

faFilter is from the UCSC source code collection, see http://genome.ucsc.edu/admin/jk-install.html or also http://genomewiki.ucsc.edu/index.php/The_source_tree

Per usual, BioPerl has the answer.

http://doc.bioperl.org/releases/bioperl-current/bioperl-live/Bio/DB/GenBank.html

# you could make an array of IDs you need to fetch
use Bio::DB::GenBank;
$gb = Bio::DB::GenBank->new();
$seq = $gb->get_Seq_by_id('MUSIGHBA1'); # Unique ID
@seqCoords=(
  [0, 100],
  [1000-1100]
);
$subseq=$seq->subseq($$seqCoords[0][0],$$seqCoords[0][1]);
# then, look at the blast modules and SearchIO to see how to start blasting and parsing
# http://www.bioperl.org/wiki/HOWTOs

You can download your genomes, build a BLAST database with formatdb and then extract a second set of sequences using fastacmd:

ncbi/build/fastacmd has a option -L

  -L  Range of sequence to extract (Format: start,stop)
      0 in 'start' refers to the beginning of the sequence
      0 in 'stop' refers to the end of the sequence [String]  Optional
    default = 0,0

then run your blastall query with the second database.

Might help: http://www.dcode.org/sequences.php