I assume you mean Devel::Cover. If you haven't already, that documentation is really good, and covers some of your questions.

If you're getting unwanted files in your output, you can use -ignore or +ignore with a regexp. But you normally shouldn't be seeing those included modules. What does your directory structure look like, and how exactly are you invoking cover?

I use Devel::Cover in my distributions. In that case, I simply run cover -test in the root of my distribution, which runs make test and reports the coverage, and won't report coverage of other external modules. Here's what the output looks like for me:

pts/6 ryan@pi:~/src/Text-Trim $> cover -test
Deleting database /home/ryan/src/Text-Trim/cover_db
cover: running make test "OPTIMIZE=-O0 -fprofile-arcs -ftest-coverage"
+ "OTHERLDFLAGS=-fprofile-arcs -ftest-coverage"
PERL_DL_NONLAZY=1 "/usr/bin/perl" "-MExtUtils::Command::MM" "-MTest::H
+arness" "-e" "undef *Test::Harness::Switches; test_harness(0, 'blib/l
+ib', 'blib/arch')" t/*.t
t/00..load.t ..... 1/? # Testing Text::Trim 1.03, Perl 5.026001, /usr/
+bin/perl
t/00..load.t ..... ok   
t/01..trim.t ..... ok     
t/02..ltrim.t .... ok     
t/03..rtrim.t .... ok     
t/04..unicode.t .. ok   
t/05..undef.t .... ok     
t/99..pod.t ...... ok   
All tests successful.
Files=7, Tests=64,  2 wallclock secs ( 0.02 usr  0.00 sys +  1.91 cusr
+  0.09 csys =  2.02 CPU)
Result: PASS
Reading database from /home/ryan/src/Text-Trim/cover_db


--------------------- ------ ------ ------ ------ ------ ------ ------
File                    stmt   bran   cond    sub    pod   time  total
--------------------- ------ ------ ------ ------ ------ ------ ------
blib/lib/Text/Trim.pm  100.0  100.0  100.0  100.0  100.0  100.0  100.0
Total                  100.0  100.0  100.0  100.0  100.0  100.0  100.0
--------------------- ------ ------ ------ ------ ------ ------ ------

HTML output written to /home/ryan/src/Text-Trim/cover_db/coverage.html
done.
[download]

I am not exactly sure what you mean by "merging", but if you can get all of your source code into the same distribution, you can have one report cover the whole thing. If that is not desirable or feasible, then I'm not sure what to suggest, aside from a simple concatenation or pagination of the results.