Strange results in the folder learningSample

I used iota2 with the mode 'constant' for sampling strategy. In fact, I wanted a maximum of 1000000 samples by class. I had 15 classes. In the folder sampleSelection, I had this file : samples_region_1_seed_0_outrates.csv which means that totalSamples*rate were used for learning. samples_region_1_seed_0_outrates.csv:

#className requiredSamples totalSamples rate
1	1000000	170747	1
10	1000000	468110	1
11	1000000	6960	1
12	1000000	508186	1
13	1000000	26932	1
14	1000000	138999	1
2	1000000	6898616	0.144957
3	1000000	72	1
4	1000000	689238	1
5	1000000	75349	1
6	1000000	155329	1
7	1000000	1056372	0.946636
8	1000000	209917	1
9	1000000	4233	1

But in the folder learningSample, I had this file: class_statistics_seed0_learn.csv

,Class,Label,Region 1,Total
1111,proteagineux,1111,0,0
1121,ble tendre h,1121,1000000,1000000
1122,orge h,1122,1000000,1000000
1123,triticale h,1123,1000000,1000000
1124,orge printemps,1124,1000000,1000000
1125,ble dur h,1125,1000000,1000000
1211,mais,1211,1000000,1000000
1221,tournesol,1221,1000000,1000000
1231,riz,1231,0,0
1251,colza,1251,1000000,1000000
1241,soja,1241,1000000,1000000
1261,tubercule racine,1261,1000000,1000000
1311,prairie,1311,1000000,1000000
1321,vergers,1321,1000000,1000000
1331,vignes,1331,1000000,1000000
0,nodata,0,0,0
Total,,,13000000,13000000

But there are not 1000000 learning samples for all this classes (blé tendre, orge...) and there is not 0 learning sample for the class 'proteagineux' because in my file T30TXT_seed_0_learn.sqlite (in the folder dataAppVal) there are samples 'proteagineux'. Moreover in the first file there are 14 classes which is normal because I have not learning sample for the class 'Riz'. But I don't think the className of 'Riz' after encoding was 15...