Some degree of "forgiveness" is not only humane, but consistent with the purported goal of kids' learning things by the time the course is over, not necessarily on the schedule (a pure artifact) of the syllabus.
One skewing is that presumably the course "builds", so that later exams presume competency in earlier... so a good early score is not at all a substitute for a good later score.
Keeping the latter in mind: in all my teaching, a good later score trumps a bad earlier one, period. Sure, it's not a good thing that people messed up earlier, but if the real goal/issue is acquisition of certain ideas, why do we care what schedule they're on?
At the same time, yes, most assuredly, a too-generous-sounding description of grading system will "confuse" many people, certainly undergrads, and often beginning grad students who've spent the previous four years (not to mention high school) figuring out how to avoid being pwned by The System... as opposed to believing (which is not at all true...) that the system has benign interest in them.
That is, even if not discussed too fully with students... who will often seriously/fatally misunderstand ... think of the genuine goals, namely, imparting some knowledge by the end of the term. That is, adherence to schedule of acquisition is really not essential, though it'd be handy.
All of this changes, of course, if by mischance the true purpose of the class is "filter/weeder", in which case one might contemplate issues unrelated to the course content, or even course structure. That is, if the filter might be to reward "steadiness" (not a bad thing at all!), then don't give a break to people who aren't "steady".
(Sure, there's hubris aplenty in figuring out or caring to speculate what the true purpose of most math courses might be...)