Assembly vs mapping results

[duonglinhnam]

Hi guys, I performed RNA-seq analysis of bacterial transcriptome in four different stressed conditions, mapping the reads on its own genome available in NCBI. Then I used FeatureCounts for reads counting and finally I performed differential analysis with NOISeq R package, because of the absence of replicates.
Before that, my tutor submitted the analysis to a famous company requiring a de novo assembly (they used the trinity pipeline for assembly and differential analysis).
I used the same fastq files, and finally I found a larger number of DE genes, but my results are the opposite of company's results. How is it possible? I know that mapping is better than assembly when a reference genome is available and above all I know that the trinity pipeline have some problems for differential analysis, because it uses DESeq or edgeR after quantification by RSEM.
What do you think about?

[GenoMax]

This question is difficult to answer without access to data. Why did the company used trinity when you have bacterial genomes and don't expect any splicing. Trinity only assembles transcripts and does not do any diff analysis. Did you not tell the company about the reference genomes available when you asked them to do sequencing?

Were you mapping reads to a very close genome at NCBI? Bacterial genomes can vary a lot even for type strains for what is in your lab and reference at NCBI.

[Bukowski]

What do you mean by 'opposite'?