We're going to make this simple.
First, of course, in order to do this, he first had to reduce the document so that the margins were the same, since the original PDF distributed by CBS is quite a bit larger. Then he superimposed the two documents, such that the margins on all sides lined up.
What he then discovered is that Times New Roman typeface is, when viewed on a computer monitor, really, really similar to Times New Roman typeface. Or rather, really really similar to a typeface that is similar to Times New Roman typeface.
Um, OK then.
You see, a "typeface" doesn't just consist of the shape of the letters. It also is a set of rules about the size of the letters in different point sizes, the width of those letters, and the spacing between them. These are all designed in as part of the font, by the designer. Since Microsoft Word was designed to include popular and very-long-used typefaces, it is hardly a surprise that those typefaces, in Microsoft Word, would look similar to, er, themselves, on a typewriter or other publishing device. That's the point of typefaces; to have a uniform look across all publishing devices. To look the same. You could use the same typeface in, for example, OpenOffice, and if it's the same font, surprise-surprise, it will look the same.
So kudos on discovering fonts, freeper guy.
Next, however: do they really match up? Well, no. They don't.
If you shrink each document to be approximately 400-500 pixels across, they do indeed look strikingly similar. But that is because you are compressing the information they contain to 400-500 pixels across. At that size, subtle differences in typeface or letter placement simply cannot be detected; the "pixels" are too big. If you compare the two documents at a larger size, the differences between them are much more striking.
For instance: In the original CBS document, some letters "float" above or below the baseline. For example, in the original document, lowercase 'e' is very frequently -- but not always -- above the baseline. Look at the word "interference", or even "me". Typewriters do this; computers don't. Granted, if you are comparing a lowercase 'e' that is only 10 or 12 pixels high with another lowercase 'e' that is only 10 or 12 pixels high, you're not going to see such subtleties. That doesn't prove the differences aren't there; it just proves you're an idiot, for making them each 12 pixels high and then saying "see, they almost match!"
"This typeface -- Times New Roman -- didn't exist in the early 1970s."
There are several problems with this theory. First, Times New Roman, as a typeface, was invented in 1931. Second, typewriters were indeed available with Times New Roman typefaces.
And third, this isn't Times New Roman, at least not the Microsoft version. It's close. But it's not a match.
For example, the '8' characters are decidedly different. The '4's, as viewable on other memos, are completely different; one has an open top, the other is closed.
So yes, we have proven that two typefaces that look similar to each other are indeed, um, similar. At least when each document is shrunk to 400-500 pixels wide... and you ignore some of the characters.
"Documents back then didn't have superscripted 'th' characters"
That one was easy. Yes, many typewriter models had shift-combinations to create 'th', 'nd', and 'rd'. This is most easily proven by looking at known-good documents in the Bush records, which indeed have superscripted 'th' characters interspersed throughout.
"This document uses proportional spacing, which didn't exist in the early 1970s."
Turns out, it did. The IBM Executive electric typewriter was manufactured in four models, A, B, C, and D, starting in 1947, and featured proportional spacing. An example of its output is here. It was an extremely popular model, and was marketed to government agencies.
"OK, fine, but no single machine had proportional spacing, 'th' characters, and a font like that one."
No, again. The IBM Executive is probably the most likely candidate for this particular memo. There is some confusion about this, so to clear up: the IBM Selectric, while very popular, did not have proportional spacing. The Selectric Composer, introduced in 1966, did, and in fact could easily have produced these memos, but it was a very expensive machine, and not likely to be used for light typing duties. The proportional-spacing Executive, on the other hand, had been produced in various configurations since the 1940's, and was quite popular.
(Note: However, it is not immediately clear that the Selectrics and Selectric IIs could not in fact emulate "proportional" spacing. There is skepticism in some circles that these memos really show "proportional" spacing. Looking at the blowups, it appears pretty obvious to me that there is, but still researching.)
http://www.dailykos.com/story/2004/9/10/34914/1603