Google truncates on width
The cut is made at a pixel measurement, not a character count. A desktop result gives a title roughly 580 pixels and a description roughly 920; a phone gives less. Nothing in that measurement knows how many letters you used.
The character rule survives because it is a decent average. Ordinary English lowercase runs about 9 to 10 pixels per character at the size Google sets a title, so sixty characters lands near 580 and the advice works. It stops working the moment the text is not ordinary.
The demonstration
Illinois libraries and MMMMMMMMM WWWWWWWW are both eighteen characters. The second is more than twice as wide. Capitals, and the wide letters in particular — M, W, @, % — spend the budget far faster than the narrow ones. A title in title case with three capitalised words is measurably longer than the same words in sentence case, and a character counter reports them identical.
This is why a brand name at the end of every title is more expensive than it looks. It is the same words on every page, priced in pixels, and it is the part that gets cut.
Arabic breaks the count outright
Harakat — the fatha, damma, kasra and sukun — do not advance the cursor. They stack above and below the letter they belong to. A vowelled Arabic word is longer as a string and identical as a rendered line.
دمج ملفات PDF بلا رفع is 21 characters. Fully vowelled it is 34 — 62% more. Measured, both come to the same 206 pixels. If you are writing Arabic titles against a character counter, you are cutting lines that were never close to the limit, for a reason that does not exist.
It runs the other way too. Arabic letters average narrower than Latin ones, so an Arabic title that a counter calls long often has room to spare — and an Arabic description usually has room for a whole extra clause that nobody writes.
Where the cut lands matters more than that it happened
A truncated title is not automatically a problem. What matters is which half survives. Cut after the brand name and you have lost nothing. Cut after the verb and the result says How to merge PDF files without… and the reader supplies an ending you did not write.
So put the thing the page is about first, and the qualifiers last. That ordering is worth more than squeezing under a limit, because it makes truncation harmless instead of making it impossible.
Leave yourself room
A line measuring 578 against 580 is not safe. Any estimate has error, Google renders in its own font on its own device, and pages grow a word at a time. Treat anything within about 5% of the budget as already over.
What fitting the budget does not buy you
Google rewrites titles it considers a poor match for the query, on a large share of results. It usually substitutes an h1 or the site name. Fitting the budget prevents your title being cut; it does not prevent it being replaced. Those are different failures and only the first is yours to control — the second is answered by making the title an honest description of the page, which is what stops Google looking for a better one.
Checking it
The SERP preview measures both lines in pixels, shows the cut where it will fall, and handles Arabic correctly. For a page that already exists, paste its source into the meta tag analyser — it measures the same way and reports the rest of the head at the same time.
Neither sends anything anywhere, which matters more here than it sounds: an unpublished title is often the most commercially sensitive line on a marketing site.