They're one byte big because, when they overlap, they only have a block in one position.

So, if you have 10001, that's normal (left side of the first character in the upper left).

Then you have 01001, then 00101, then 00011, then 00001.

Also, all dot positions are not valid; one is always at the top or bottom (and in the case of the first column and first row, top and bottom).

Having stuff in the center columns doesn't impede machine reading for two reasons:

1. when they're irrelevant, machines don't bother to read the center column; they just read the edges (like a human reader would, anyway). 
2. Academic image-processing software is very, very good. It might not get the expression of the line, or get that you drew this word shaped like a penis (they're human; they think penises are funny, too), but the software knows to look at the edges.