Sure,

You always have these alternatives, but the problem is that it starts to
become harder to restore your wallet from the initial mnemonic if
something goes wrong.

Say you lose all your wallet information except for your mnemonic,
extended public keys from all people you established multi-signature
accounts with, and you know which arrangement you had with each of your
cosigners (2-of-3, 2-of-2, etc.).  Your software will not have a hard
time rebuilding all accounts from information obtained from the public
blockchain.  Adding a new dimension, here the i' derivation, will make
things harder.  You would need to know this piece of data too.

Another good thing about using the same derivation always is that you
can give every cosigner only that single piece of information, that
single xpub, in order to establish multisig relationships.  There is no
need to use a different one per relationship.  This simplifies the
workflow for establishing new multi-signature accounts significantly.

Greetings

On 05/10/15 13:32, Jonas Schnelli wrote:
> What holds you back from using m/i'/45' where i' is your multisig
> "account" number?
> 
> Within your BIP45 wallet (lets assume Copay), you would not provide
> the xpubkey at m/45', instead you would provide your xpubkey at m/i'/45'
> .
> 
> It's probably no longer pure BIP45.