I have another tack on this problem, it might not be better though. If the number of fields is consistent and the only field that contains commas is the company name you could rebuild the company name after splitting. The following seems to work.

use strict;
use warnings;

my $max = 4;  #Largest expected index.

while (<DATA>) {
    s/#.*$//;  #Clear out the comments
    chomp;
    my @foo = split /,/;
    my $company = $foo[1]; #prime a variable with a company name
    if ($#foo > $max) {        #If there are more elements than the ma
+x...
        my $range = (scalar @foo) - $max; #Get the top of the range
        $company = join '+', @foo[1..$range]; #The company name is in 
+a slice
    }
    print "$company\n";
    #do whatever with the data.
}

__END__
client,ABC Company, jwilkens,jdoe,mdeart #this one is fine
client,Cornell,Thayil,Cameron & Sheppard, LLC, klivingston,amarison,ps
+erton #thi
s one is not
client,Doe,Smith & Randall, Inc, jwaters,pfloyd,jjoplin #this one is n
+ot
[download]

I didn't bench this against substr and index, that might be more efficient, and I now this code could be compressed a bit but I thought this was a fairly clear way to write a sample.

Ira,

"So... What do all these little arrows mean?"
~unknown