The cryptogram
Transcribed from the 1974 facsimile and checked digit by digit against the page image, then against Schmeh's published text.
The zero at digit 195
Every digit in an odd position is 6 7 8 9 and every digit in an even position is 1 2 3 4 5 — across all 392 working digits, with exactly one exception. Digit 195, the last digit of group 39 (63630), is a 0 sitting where a row digit belongs.
The page was re-rendered at 1400 dpi to settle it: the glyph is an old-style zero, an x-height oval, unmistakably not a 6 or an 8. It is real, and E3 below argues it is not an error either — read the row digit as the row number plus five and the labels run 6 7 8 9 10, with 10 set as a single figure. The zero is row five of the square.
And it does not sit just anywhere. It is unit 98 of 196 — the exact midpoint of the message, and the last cell of row 7 in the 14 × 14 grid. That may be coincidence in a message with only one anomaly to place, but it is the sort of coincidence that deserves testing rather than admiring, and it is item 3 in Section 8.
The Workbench lets you overrule the whole reading and substitute 6, 7, 8 or 9 instead. Every replacement has been run through the full machine attack; none changes the outcome.
The odd digit count
395 digits will not divide into pairs. Drop the trailing 000 of the final group as transmission padding — the convention d'Agapeyeff uses throughout the book, "nulls to complete the five letters" — and 392 remain.
392 digits = 196 units, and 196 = 14 × 14. A perfect square, which is exactly the shape the book tells you to look for when attacking a transposition: "we count the number of letters in the crypt; there are 36, which is 6 × 6, and we therefore construct a square accordingly."
Tempting, and probably deliberate. Section 4 reports what happened when every transposition that shape allows was actually tried.