If by "small" you mean less than 100nts, then obviously you have lost them during sequencing; otherwise, it depends on the package you use to map your reads into transcripts, and whether you have mapped them based on known transcripts only or with an option to find novel splice variants as well.

how exactly did you do mRNA enrichment? if it was poly-A enrichment etc, chances are you have lost most, if not all of the small RNAs in that step itself. Also, I'd think that most methods of library construction for mRNA sequencing throw away very small sizes right at the beginning and most of the small RNAs that you might be interested in (i.e. miRNAs etc) are just too small to have been enriched in that library. Usually, library preparation and sequencing etc are all fairly optimized for the final purpose and the same data can't be used for detecting both mRNAs and small RNAs. 

I forgot to mention that was bacterial RNA. Obviously small RNA were lost during sequencing since 100 nt were sequenced from each side. I concentrated my mind only on RNA protocol and therefore my question was little limited. Thank you both.

I found a paper where they compare number of sRNA when different libraries are used, Lib>100; Lib

Dear Loubna,

you can easily check for your sRNAs, even if your extraction method was focussing on longer RNAs. We find regularly sRNA in longRNA deep seq experiments -if they are robustly expressed, that is. However, if you want to publish the data I would recommend repeating the experiment with a sRNA deep seq protocol. 

One thing to keep in mind if you want to find small RNAs is the way you extract the RNA. Most column-based kits for total RNA will not efficiently retain RNAs smaller than 100-200nt. Strongly transcribed RNAs like tRNAs will still be present, but I would not recommend such a kit for that kind of application.

BTW: How, i.e. with which program, are you looking for the small RNAs?