Can not determine the ARIMA model for data 

I am trying to develop a useful statistical time series model (ARIMA) using auto arima. The program reported that it can not determine the model. Please help me understand the limitations (if any) of using this software on my data.
Here is my data (26 values monthly):
23,887 14,914 7,751 1,270 4,836 11,349 18,480 21,305 11,882 9,797 9,122 9,965 5,906 12,025 16,012 14,683 21,876 69,097 68,737 78,631 125,726 128,841 169,854 333,288 76,395 69,686 
 A: There are many kinds of transformations many with very bad side-effects as @mpkitas pointed out like taking logs of this data set. The 26 historical values suggest 3 unusual values (24,23 and 21 in order of importance ) thus one fairly non-intrusive transformation is to adjust the observed values for these three periods. A potentially usable model (nearly a random-walk with ar(1) of .937 developed by apiece of software that I have helped develop is then  . The Actual/Fit/Forecast graph using classic assumptions regarding the estimated parameters being identical to the population parameters and no unusual values in the future is here  . If one allows uncertainties in the model parameters to come into play we get slightly wider confidence limits  . Finally if we also add the possibility of anomalies arising in the future (historically nearly 10% i.e. 3 out of 26) we get a more realistic ( somewhat daunting !) picture .
In my opinion data sets like this arise and simple textbook approaches just don't cut it. Bacon pointed pointed out  and in this case the data is suggesting that our understanding of the application of good statistical reasoning (ARIMA) needs to improved incorporating more robust approaches as shown here.
