For anyone to offer 25 selections is inevitably going to reflect personal
bias.  The method depends on your reasons for such a list.  If your reason
is to boil down the whole of "classical" music from earliest to the
present, then one method would be as follows.

1.  For each century, (pick an arbitrary starting point.  For practical
purposes, I suggest the 10th) determine the most-used genres (in terms
of their representation in libraries, performance records, contemporary
accounts, etc) in both vocal and instrumental (maximum of 5 per century?)

2.  For each genre, determine a small number of most-performed or
most-published works.

3.  From the list in No.  2, rank the composers (including "Anonymus")
in one of two ways depending on your emphasis:  1) most widely performed
in their time, or 2) most highly regarded today (number of recordings,
performances, etc.  in catalog, number of columns in music encyclopedias,
etc).

4.  Calculate a "century skew factor" based on number of available
recordings in catalog from each century, and/or sales of recordings from
particular centuries.  (I'm guessing this will still be shaped like a two
humped camel--one Baroque and one 19th-century) and assign choices on a pro
rata basis.

5.  Do not be concerned with deciding if a particular work is "pivotal"
or "essential," especially if your goal is to understand how music was
perceived at the time the work was written.  Otherwise you run the risk
of taking "conventional wisdom" to be synonymous with true importance.
It may be, but it's wise to be skeptical.

If more than 25 works rise to the top, pare down the list making genres
the last to go.

Who knows, you may wind up with something by Clementi or Telemann!

That, I repeat, is only ONE way to approach the task systematically.
The aim of any such exercise should be to come up with a list that can
be justified on some rational basis, rather than on personal likes and
dislikes.

Chris Bonds