It is important to note for those of you finding this issue that the extra classes are a result of input values not being accounted for in the input reclassification table (remap). Whether entered manually, provided by default, or through the available tool classification methods.
For the case defined in this post, the input raster is a floating point raster with many unique values. The method of creating the remap was based on an estimated histogram, which calculated a minumum and maximum not equivalent to the actual range of values in the input raster. Therefore, not all input values were correctly classified, and by default, missing values in the reclass table retain their input value (the option is to map them to NoData), hence the extra output values.
To correctly classify your data your options are to,
1) manually edit your reclassifcation table range of values and provide the correct break values (including the actual maximum).
2) use Windows registry setting to calculate the histogram using all unique values (as mentioned in the KB article by Eric). Please note this may take extra time depending on the size of your dataset.
3) if the desired output is a sequential range of values (ie, 1-9) you can try the Slice tool.
We will work to find a way to provide you with the most accurate reclassification table by default, in the mean time hope this helps you get your work done.
Regards,
-Ryan
