xbee questions (baud rate, transparent mode error correction)

[linux-works]

I've been using xbees for a while but some of the details are still unknown to me.

for example, if I'm using simple 'non-api' mode and just sending in serial ttl data to one xbee and getting that exact stream out of the other (what I would call 'transparent mode'), I notice that I don't get an error free transport.  when the signal is good or if I use xbees with antennas soldered to them (factory), what I see is what I put into the other end.  when I use the chip-antenna non-wire version or if the signal is not strong enough I get dropped chars and the stream is not truly trustable.

is there a setting to ensure, even with retries done for you, that you get an error free end to end link in 'simple' non-api mode?

if I add a bunch of other xbees all tuned to the same network id (ATID 3331, for example) does the thruput or error detection/correction get worse?  I'm using the style where one transmits and the rest all receive.  I don't know if this incurs overhead or if its 'free'.  anyone have experience in this area?

other question is about baud rate, still using simple non-api mode.  do the end points have to be using the same baud rate or is that just a local issue and not relevant to the rf to rf link between the radio pairs?  suppose we ignore buffering issues and how long it takes to empty a buffer at slow serial speed vs higher speed.  suppose the packet is a few characters, very very small.  would either end be free to set its local baud rate to what it wants or will there be problems if they are mismatched?

(I have the series-1 xbees, not the pro or series-2).