check if encrypting the case-flipped letter (since xor-ing with a white space flips the case of a letter) with k_sub yields c_sub2, if it does we should know that k_sub is the correct sub-part of the key. If not, try the same with assuming that m_sub2 is a whitespace
Currently you are pairing two ciphertext streams, while you have 10 of them. You should choose one stream out of 10 and then perform the XOR with all other 9 at the same position. If most of those streams return letters (and no invalid combinations of a XOR - but that's harder to test) then you can be pretty certain that the encrypted character is a space, and you can find the key by the XOR you perform in your answer.
I think it might be the case in Boneh's assignment, but beware that you may encounter plaintext combinations that together may produce a letter as well. So you cannot simply perform a XOR with two streams, and validate your guess that way. The more streams you have, the more sure you are; in the end you can of course validate by looking at the plaintext messages as well and/or use your language skills to fill in the blanks.
If I remember correctly I took it on a bit differently. I took one stream, iterated over all characters in the alphabet (note: the possible input characters is called "the alphabet" here, I don't mean the ABC) for each position, and then XOR'ed all three together. If all streams produced results in the alphabet then I hit the right character. This allows you to only handle characters in the alphabet, not weird XOR combinations.
If multiple characters are found, then use the one that produces the most used characters in all the streams together (frequency analysis).
If that still doesn't produce results you can try other streams as well (you would only have to compare stream 2 with 8 streams of course, as you already compared #1 and #2).
OK, so lets say we have the first position (positions are not indicated in the variable names) and a set of streams. Now let us start with the first stream, and XOR it with all other streams, denoted by $y$ where $y != 1$.
You have a pair of two ciphertext values that consist of $c_1 = p_1 \oplus k$ and $c_y = p_y \oplus k$. So XOR'ing together gives you $r = c_1 \oplus c_y = p_1 \oplus p_y$ (nothing new here). Now if you guess $p_1$ and call it $p'_1$ you would get $p'_1 \oplus p_1 \oplus p_y = p'_y$. Now if $p'_y$ is an invalid character then obviously $p'_1$ was a wrong guess. If you're unlucky then you need to perform frequency analysis on all the resulting $p'_y$. But remember that you can do this with all $\binom{n+1}2$ pairs before resorting to that.
Once you have $p'_1 = p_1$ then obviously the key is just the XOR with the ciphertext character: $c_1 \oplus p_1 = k$. That means that you only have to iterate over the alphabet rather than all the keys, and you can really quickly dismiss bad tries.
