Docstring for urlunparse says:
    """Put a parsed URI back together again.  This may result in a
    slightly different, but equivalent URI, if the URI that was parsed
    originally had redundant delimiters, e.g. a ? with an empty query
    (the draft states that these are equivalent)."""

“Draft” here refers to RFC 1808, superseded by 3986. However, RFC 3986
(section 6.2.3) states:
“Normalization should not remove delimiters when their associated
component is empty unless licensed to do so by the scheme  
specification.  For example, the URI "http://example.com/?" cannot be  
 assumed to be equivalent to any of the examples above.  Likewise, the 
  presence or absence of delimiters within a userinfo subcomponent is  
 usually significant to its interpretation.  The fragment component is 
  not subject to any scheme-based normalization; thus, two URIs that   
differ only by the suffix "#" are considered different regardless of   
the scheme.”

I guess we need some tests here to check compliance.

This is indeed a bug. urlunparse should special-case "#" so as not to
discard it.

Currently this claim will fail:

>>> obj = urlparse.urlparse('http://a/b/c?')
>>> urlparse.urlunparse(obj)
'http://a/b/c'
>>> obj = urlparse.urlparse('http://a/b/c#')
>>> urlparse.urlunparse(obj)
'http://a/b/c'

If we move away from the current behavior, there will surely be some test failures that can be observed for urljoins. We will have to consider those cases too while fixing this.