According to CNN, ballots are still being counted across the USA (as of 11/16/16). While in many states, the number of counted ballots is too small to change the result, the outcome in Michigan is labelled as unknown and for instance Arizona has only reported 80% of the votes and the gap between the two main candidates is only 4.33% of the total number of votes. A similar situation is observable in Wisconsin (95% reported and a gap of less than 1%). Suppose MI, AZ and WI go to Clinton which is, although unlikely, quite possible, that would give a 269-269 electoral college deadlock. So why did Mrs Clinton concede so early ?
There are not really any downsides to Clinton conceding early. As noted in this related question, Al Gore conceded and then later withdrew the concession and decided to fight the court case. So, once you compare the positives and negatives, it becomes clear that Clinton conceding actually benefits her quite a bit even if the election results are later declared tied or in her favor.
- She looks like a gracious loser, which will help in other runs for public office
- She gets to make a concession speech where she can try to rally her supporters, again supporting a run for another public office
- Whether she cares or not, she can claim that she did it for the better of the country, since it allows the President Elect to start the transition process, rather than being stalled and having to wait for the results to be finalized
- Some might see it as weak, or think that she thew in the towel early here and will do it in office
I put in the negative as a possibility, but I honestly do not think it is a significant concern, and will certainly be outweighed by the negative press of her trying to fight an election that clearly looks hopeless for her at this point. In the situation that the election does turn around, I think a very insignificant number of voters will fault her for having conceded when winning looked impossible and her concession meant a smoother transition into the next administration.
Additionally to IllusiveBrian's answer, you should consider what these numbers mean.
While it is mathematically possible that any of those close races changes sides, it is very difficult as it would mean that the remaining votes do not follow the same pattern as the already counted ones, and by a considerable margin.
For example, from your data, for Winsconsin (95% reported, 1% gap). So, for each 100,000 votes, we have:
95,000 counted, 5,000 yet to be counted.
Of the 95,000, 50.5% go for Trump (47,975) and 49.5% go for Clinton (47,025). Yes I am ignoring third parties here, it is just a thought experiment. A 950 difference for Trump, so far it does not seem to be that difficult.
In order of upsetting that, you need that the remaining 5,000 distribute at least as 2,975 for Clinton and 2,025 for Trump. Which means that Clinton needs to win a 59,5% of the remaining votes... ten full points over the pattern of the votes already counted!
Normally this kind of changes would only be realistic if there were other correlations (e.g., most of the already counted votes are from rural areas and the votes remaining are mostly for cities, etc.) that could explain such a change in pattern.
It is the same principle than polls, but with a way bigger sample (which means less error is to be expected)
Add into the calculus that you would need all three of the states to change to break Trump's majority, and the probabilities go to barely zero (if there is, a 10% probability1 of any single state switching, the probability all of three switching is of 0.1%2).
That is why many people may accept as good the electoral results even when the number of votes not computed is still greater than the difference between the candidates.
And then, of course, the answer by IllusiveBrian (conceding defeat does not change the results of the election and does not compromise Hillary's options in the very improbable case of a different results) and Count Iblis comment (if there is a tie then the POTUS will be decided by the Republican controlled House of Representatives).
1Which I think is a very generous estimate, but happens to be easy for the calculations.
210% = 0.1 probability. The probability of 3 independent events is the multiplication of their probabilities, so for the three states changing their winner it would be 0.1 * 0.1 * 0.1 = 0.001 = 0.1%