The switches -lanF: basically add some code to the oneliner provided with -e, this is described in perlrun, but you can also use B::Deparse to turn the oneliner into a script to see it:

$ perl -MO=Deparse -F: -lane 'push @x, $F[0] if /^[A-Z]/}{print for so
+rt @x'

BEGIN { $/ = "\n"; $\ = "\n"; }
LINE: while (defined($_ = readline ARGV)) {
    chomp $_;
    our @F = split(/:/, $_, 0);
    push @x, $F[0] if /^[A-Z]/;
}
{
    print $_ foreach (sort @x);
}
[download]

Which shows you that the script is:

1. Setting the input record separator $/ and the output record separator $\ to a newline "\n" (the -l switch), in this case this is mainly useful for not having to add a newline when printing output
2. Reading the input line-by-line in a while loop (the -n switch), and for each line:
   1. Removing the input record separator (newline) with chomp (the -l switch)
   2. splitting the line on the character : into the array @F (the -aF: switches)
   3. pushing the first element of the array, $F[0], onto the array @x, if the input line $_ matches /^[A-Z]/ (the code from the -e switch)
3. After that loop,
4. Loop over the sorted elements of @x, printing each one. (also the -e switch, after the }{)

Cleaning up the above code gives you this script:

use warnings;
use strict;
my @out;
while (<>) {
    chomp;
    my @fields = split /:/;
    push @out, $fields[0] if /^[A-Z]/;
}
print "$_\n" for sort @out;
[download]

Update: Added references to the corresponding switches to the above list and expanded explanation of -l.