It simply expects a timezone:

>>> from pywikibot.textlib import TimeStripper
>>> import pywikibot
>>> site = pywikibot.Site('commons', 'commons')
>>> ts = TimeStripper(site)
>>> ts.timestripper('20:43, 2 July 2017')
>>> ts.timestripper('20:43, 2 July 2017 (UTC)')
Timestamp(2017, 7, 2, 20, 43, tzinfo=tzoneFixedOffset(0, UTC))
>>> ts.timestripper('{{20:43, 2 July 2017 (UTC)}}')
Timestamp(2017, 7, 2, 20, 43, tzinfo=tzoneFixedOffset(0, UTC))
>>> ts.timestripper('{{20:43, 2 July 2017}}')

In T170034#3417683, @zhuyifei1999 wrote:

It simply expects a timezone:

Yes, but in this example SignBot uses {{unsigned2|18:32, 5 July 2017|Mrspeedybug}} that prints (UTC) by default.
Do we need to code an exception for this template on Commons ?

IMO, no.
Each wiki might use a different (set of) templates.
Some other workaround should be found.
I will see if I can figure something out.

In T170034#3419143, @Mpaa wrote:

Each wiki might use a different (set of) templates.

When I wrote this i was thinking about something like the list of per-site category_redirect_templatesthat exist in pywikibot/families/wikipedia_family.py

Not sure if we want to remove the timezone requirement from parsing. We want to discriminate timestamps from other mentions of date and time. An explicitly written timezone symbol is an indication that it's probably not a part of a normal sentence.

TimeStripper is (sometimes too much) flexible and accepts a few characters in between. E.g.

>>> ts.timestripper('We will meet on 1 Jan 2018 at 12:30 (UTC)')
Timestamp(2018, 1, 1, 12, 30, tzinfo=tzoneFixedOffset(0, UTC))

The parsing itself has been a bit heuristic (there are at lease some false positives), but if we simply drop the timezone requirement, I fear that false positives increase. So if we are to do it, it will have to be dropped only for a few selected templates.

I do not think we should remove timezone from parsing.
We could not do Timestamp -now() reliably.