Be sure to share the concern expressed by others about computing
statistics on so-called "qualitative" data.  I have mixed feelings in
this case.

In light of those concerns, my explanation below should be taken with
a grain of salt.  It is intended to be general in nature.

If you want to compute statistics that describe just those 30
respondents, they are "population" -- that is, the complete set of
data that you want to talk about.  Given the dubious use of statistics
on movie rankings in the first place, I would say this is your own
option.

On the other hand, if you want to compute statistics about the 30
responses for the purpose of drawing inferences about a larger
population, the 30 responses represent a "sample" of the larger
population.  That calls for different formulas -- usually minor
differences like subtracting one here and there.  For example, use
Help to see the difference between STDEV and STDEVP.

The concern that others have expressed is:  usually we can compute
statistics like standard deviation only on quantitative variables --
variables that represent a count or measurement.  Movie rankings are
usually considered qualitative data -- like color of the hair.  For
qualitative data, we usually use percentile statistics and visual
methods of comparison.

(Example:  If 10 people have blond hair and 5 people have brown hair,
you cannot say that the average hair color is "blond and a third",
even if you assign 1 to blond and 2 to brown.  Many people the same
thing about movie rankings.)

First, I never said it would be useful or "logical".  I was simply
giving you what you asked for:  the Excel formula to compute the
mathematical definition of std dev for grouped data.

Second, your conclusion ("values lie +/- 0.8 units around 4.5") is not
logical insofar as that is not what the std dev tells you.  The std
dev is merely a measure of dispersion of data around the mean.  It is
not the only measure of dispersion to use in a statistical analysis,
as Mike and Stan tried to explain.  You may have simply learned that
they were right in the first place.

Moreover, the std dev does not (always) tell you anything about the
range of the data.  You might be thinking of the interpretation of the
std dev in a "normal distribution" of data, which means that the
(population) data meets certain criteria.  (Yours do not.)  But even
in that context, we would expect only about 68% of the (population)
data, not all of the (population) data, to lie within +/- 1 sd (0.76
in your case) of the mean.

Finally, even if the std dev could be interpreted as a range for the
data, the range would be clipped (bracketed) by any real-world
constraints, e.g. 5 in your case.  So hypothetically (and incorrectly
in your case), we would conclude that the range of data is -0.76 to
+0.50 around the mean.  (But I reiterate:  that conclusion is not
valid.)

At this point, I think it is fair to say that we have gone beyond the
Excel question and wandered into tutorial statistics.  I am not
prepared to explain the latter in this forum.