Actually this can't be the case. If the sample rate would be 3 bit per character there is no way to have 7 '0' in a row. With 3 bit per character there can only be '0' strings of length 3 (between dots and dashes), 33=9 (between characters) and 37=21 (between words).

Because of this the bits per character must be 1, as this is the only way we can get '111' and '0000000' at the same time

Yes there need to be at least one zero between each dot and dash, but with '1110000000111' we only have 2 dashes, each building a word of 1 character.

'1110000000111' -> '- -' -> 'T T'

We would need to have '111000111' in order to get '..' -> 'I'

'111000000000111' would lead to '. .' -> 'E E'

No. According to the problem statement there needs to be at least one zero between each dot and dash. So '1110000000111' should be decoded as '..' because the number of bit samples for a dot seems to 3 so this bit string should be decoded as 'I'.

actually "T[space]T"