Your script is flawed in two parts:

* it doesn't filter the return value of listdir(), which leads to each
test being executed twice in the presence of .pyc files

* __import__("foo.bar") does *not* return the foo.bar submodule, but the
foo module.

But that is all not relevant because your run_tests is not even called
if unittest finds command line arguments that give test names. See
unittest.TestProgram.parseArgs.