Especially the punctuation marks and most especially the period, which sometimes punches a hole right through the paper. I have a very old Underwood portable I sometimes use and from time to time I have to adjust it for this and other reasons.
Also, as the typewriter ribbon gets worn, the print tone lightens, especially for those passages typed when the ribbon is near the end.
There are lots of other peculiarities too. For instance, in an old typewriter like mine, the metal characters wear down with age and do so unevenly. This changes their appearance in ways not casually apparent but which become obvious under close inspection.
Others have noted that the wordwrap of a manual typewriter works differently than a word processor.
To analyze this facet more specifically, MSWORD wraps the current word when the end of a word passes the right margin, whereas a person typing manually finishes the word he is typing when the right-margin bell rings, and then does a carriage return. The net effect of the manual typewritten style is that, if one draws a vertical line at the bell-margin, that line should go through (intersect) the last word of every line of text.
Clearly, from the 1 AUG 1972 memo, NO SUCH VERTICAL LINE CAN BE CONSTRUCTED THAT INTERSECTS THE LAST WORD OF EVERY WRAPPED LINE OF TEXT!
For example, in paragraph 3, on a typewriter, we have to assume the right-margin bell rang while typing the word "and" on the 1st line. Yet, if that was the case, then on paragraph 2, the margin bell would have rang while typing the word "orders", and the subsequent word "for" should have appeared on the next, not the same, line.
On another thread, bolobaby noted that "I believe the documents are forgeries, but this is not definitive proof by any stretch of the imagination."
To this, I added that it may not seem definitive, until you see that the actual memo wordwrap exactly coincides with what MSWORD produces. What are the odds that a manual typist, over the span of 4 memos, would exactly choose to squeeze in the EXACT SAME words, so as to "accidentally" coincide with what MSWORD would produce. I would say, with metaphysical certitude that the odds are nil.