I would be tempted to mung the data to make the field seperator consistant and then parse it much as others have suggested-

while (<THATFILE>){
    s/^(\d+): /$1,/;    # convert first field seperator from ': ' to '
+,'
    my @fields = split /,/;
    next unless $fields[3] == 987;
    my ($hr, $min, $sec) = split /:/, $fields[1];
        # do something with time here
}
[download]

I think that makes the code clearer and I doubt any-one would look at the source data and miscount the fields, the comment would be mandatory however.

my $chainsaw = 'Perl';