For a hash table, the general "promise" is that you map a very large space of keys (e.g. all strings) to a very small space of indexes using a simple function. Common sense might tell you that this shouldn't be possible ("how can you put something big into something small?") - and of course actual hash tables solve the conundrum by acknowledging collisions and defining ways to deal with them.
But by saying "let's assume no collisions" the reader is asked to just ignore the logical contradiction instead of solving it. That can lead to frustration in the best case or to serious misconceptions in the worst case.
Of course there is nothing wrong with saying "let's ignore those points for now, I'll explain later". Also I guess it depends very much on the learning style of the reader. But bad simplification can be very annoying for some learners.
