use Perl

This means that something that matches '"quite a long phrase"' scores lower than '+word other'. So your scoring is totally different for queries according to how many distinct tokens the query contains.

The way around this is to calculate a max score for a query.. this snippet of code is quite handy for this :

#/usr/bin/perl -w

print "\nstarting...\n";
my @strings = ('"quite long phrase"','+must optional','"short phrase" word', '"quite long phrase" word');
foreach (@strings) {
    my $max = 0;
    print "string:$_\n";
    my @tokens = m/(\"[\s\S]+\"|\S+)/g;
    print "tokens:\n";
    print join(":",@tokens), "\n\n";
    foreach (@tokens) {
    $max += (m/[\"\+]/) ? 0.8 : 0.3;
    }
    print "max score : $max\n";
}

print "done...\n";