Instead of going through the string(s) one character at a time, you could split the second string into whitespace and non-whitespace parts, and then apply them blockwise. That way you have far fewer iterations and operations to do.

use 5.014;
use strict;
use warnings;

my $str1 = "12345 ABC 987 MNO";
my $str2 = "      CDE";

my $pos = 0;
my $res = '';

for my $chunk (split /([^ ]+)/, $str2) {
    if (substr($chunk, 0, 1) eq ' ') {
        $res .= substr $str1, $pos, length $chunk;
    }
    else {
        $res .= $chunk;
    }
}
continue {
    $pos += length $chunk;
}

if ($pos < length $str1) {
    $res .= substr $str1, $pos;
}

say $res;
[download]

If you really want it to be fast and correct, provide us some test cases (Test::Simple or Test::More to the rescue!) and some Benchmarks into which we can plug our solutions.