your request is hazy, seeker.
reading the title, and with heavy use of PSI::ESP qw(kreskin), I imagine your question lies specifically in making these three statements faster/better/easier :

    qr/([^$wc]+)/o,
    qr/^$ignoref([$wc]+?)$ignorel$/io,
    qr/^$match_string$/,
[download]

if this isn't the case, tighten up your question and ask again. Be specific and concentrate on your petition.

here's a regexp free implementation, in test form.

use strict;
my @wordchars = ("a".."z","#*");
my @ignorefirst = ("#");
my $test=" foo#&#bar&#baz* ";
my $wordchars=join "",@wordchars;
my $ignorefirst=join "",@ignorefirst;
#
#   get words
#
eval ("\$test=~tr/$wordchars/\\n/c;");
print $test;
my @words=split /\n/,$test;
#
#   now that the words are obtained, work on removing any ignorefirst
#
foreach (@words) {
    eval ("substr (\$_,0,1)=~ tr/$ignorefirst//d");
    # ignorelast would be similar.
}
print "search terms are : ";
foreach (@words) {print "$_ ";}
[download]

I suppose there's a performance hit for the 2 evals, but I don't know how they'd compare to your methods. I imagine you could change the relevant lines to something like

substr (\$_,0,1)=~ s/[$ignorefirst]//;
[download]

but that's untested, but I imagine (really imagining, I've no idea) it'd be pretty optimized by the regex engine since it's but one char.