You need a script for that, here my solution in Perl:

#!/usr/bin/perl

use strict;
use warnings;

$ARGV[1] or die "use substract.pl FILE1 FILE2 > OUTPUT\n";

my $file1 = shift @ARGV;
my $file2 = shift @ARGV;
my ($id, $ini, $end);

# load data from file2
my %data2 = ();
open (my $f2, "<", "$file2") or die "cannot read $file2\n";
while (<$f2>) {
    chomp;
    ($id, $ini, $end) = split (/\s+/, $_);
    $data2{$id}{'ini'} = $ini;
    $data2{$id}{'end'} = $end;
}
close $f2;

# read and process file1
open (my $f1, "<", "$file1") or die "cannot read $file1\n";
while (<$f1>) {
    chomp;
    ($id, $ini, $end) = split (/\s+/, $_);
    next unless (defined $data2{$id});
    print "$_\n" if ( $ini <= $data2{$id}{'ini'} or $end <= $data2{$id}{'end'});
}
close $f1;

Using it:

$ perl substract.pl file1.txt file2.txt
SpoScf_15890    12  2376
SpoScf_07684    10  4075
SpoScf_07684    64  4276
Super_scaffold_60   74732   78743
Super_scaffold_60   74955   77543

Here's Python script:

import sys

d = {}
for argv in [2, 1]:
    with open(sys.argv[argv]) as fh:
        for line in fh:
            sl = line.split()
            id = sl[0]
            st = int(sl[1])
            en = int(sl[2])
            if argv == 2:
                d[id] = [st, en]
            else:
                if id in d and (st <= d[id][0] or en <= d[id][1]):
                    print(line.strip())

Run it:

python substract.py file1.txt file2.txt

Output:

SpoScf_15890    12  2376
SpoScf_07684    10  4075
SpoScf_07684    64  4276
Super_scaffold_60   74732   78743
Super_scaffold_60   74955   77543

It doesn't exactly answer your question, but if you're dealing with regions you may want to have a look to bedtools:

$ bedtools intersect -a file1.bed -b file2.bed -wa
SpoScf_15890    12      2376
SpoScf_07684    10      4075
SpoScf_07684    64      4276
Super_scaffold_60       74732   78743
Super_scaffold_60       74955   77543

As I said, this doesn't answer your question since this tool looks for overlaps rather than positions-less-than conditions, but it does allow you to think about a faster and more eficient way of handling region information. Having said this, and although it's not the most efficient way to answer your second question with the "but not more than 1000 (1K)" condition, you can still solve it with bedtools:

$ perl -lane 'foreach $i (1,2) { print join "\t", $F[0], $F[$i]-1, $F[$i], (join " ", @F) }' file1.bed > file1.bed.temp
$ perl -lane 'foreach $i (1,2) { if ($F[$i] > 1000) { $t = $F[$i] - 1001 } else { $t = 0 } print join "\t", $F[0], $t, $F[$i] }' file2.bed > file2.bed.temp
$ bedtools intersect -a file1.bed.temp -b file2.bed.temp -wa | cut -f4 | sort | uniq -c | perl -lane 'print join "\t", @F[1..3] if $F[0] == 2'
SpoScf_07684    10      4075

Thanks to all,

Unfortunately, both a.zielezinski and Amigo answers could not answer my question when I added a new condition ("but not more than 1000"). As Amigo said, bedtools doesn't answer my question because this tool looks for overlaps, not positions-less-than or positions-greater-than conditions. Let me bring a new input example files and the output file which I expect to get.

file 1:

SpoScf_15890    12  1376
SpoScf_15890    12  2376
Super_scaffold_60   73836   77943
Super_scaffold_60   73926   76843
Super_scaffold_60   78936   87943

file 2:

SpoScf_15890    1   3376
Super_scaffold_60   74936   77943

output:

SpoScf_15890    12  2376
Super_scaffold_60   73836   78943
Super_scaffold_60   73926   76843

in file 1, first-line (SpoScf_15890 12 1376) will not be extracted because column 3 (1376 ) is less than 3376 but more than 1000 (1376 - 3376 = -2000 (should not be more than -1000 if want to be extracted).