It seems to me like all of your work so far has been focused solely on trying to find a common difference between the terms, to which end you've discovered there doesn't appear to be any such pattern. There's nothing wrong with this in and of itself, but it is very very limited thinking, and so it's no wonder you haven't found an answer. You need to broaden your horizons and "think outside the box," as it were. Sequences defined by common differences, called arithmetic sequences, are just one type of sequence, but there are countless others. Another common type of sequence is a geometric sequence, where each term is a constant multiple times the previous. Let's see if that holds here...
3 is the first term, so that's our base. 4 = 1.33... * 3, and 3 = 0.75 * 4. After that, we're back to 4 = 1.33... * 3. Looks good so far, maybe this is the pattern? Oh, but then 5 = 1.25 * 4, and 3 = 5 * 0.6. No dice there. So we can conclude that the sequence isn't geometric either.
This type of trial-and-error method is a great starting out method, if nothing else because it quickly eliminates types of sequence it's not and gives you guidance on how best to hone your thinking. For this particular sequence, one possible answer immediately jumps out to me just from looking at subsequences. Take only the first two terms (3 and 4). What do you notice about them? Now take only the next three terms (3, 4, and 5). What do you notice about them? Now let's look at the final three given terms (3, 4, and 5 again). What do you notice about them? Are you noticing a meta-level pattern within these subsequences? What number should come next, to continue this pattern? Finally, can you generate a subsequence of five numbers that again continues the pattern? Does this fit with the fact that 7 is given as the 14th member of the overall sequence?