To see what exactly the FDA does when evaluating a new genetically engineered crop I looked at the report about a new herbicide-tolerant soybean. The specific example is pretty much random, I simply chose to first one in the list.
They do check the sequence of the inserted gene:
Syngenta and BCS characterized the insert in SYHT0H2 soybean using nucleotide sequencing and Southern blot analysis of restriction enzyme digested genomic DNA.
They also check if the inserted gene leads to any genetic instability:
Syngenta and BCS determined the genomic stability and inheritance pattern of the inserted DNA. Genomic stability was demonstrated using Southern blot analysis of genomic DNA, which showed consistent hybridization patterns over three generations.
And they check if the inserted gene disrupts any other host genes:
To determine the site of genome insertion, the developers first sequenced the genomic regions flanking the insert in SYHT0H2 soybean and then used the results to sequence the corresponding region in the untransformed variety “Jack.” They searched publicly available databases for known coding sequences similar to the flanking sequences. Syngenta and BCS conclude that the insertion does not disrupt known soybean genes.
On the protein level they do check how much of the protein is produced:
Syngenta and BCS conducted field studies to measure expression levels of AvHPPD-03 and PAT in SYHT0H2 soybean. Tissue samples were collected from SYHT0H2 soybean and control soybean grown at four locations in Argentina during the 2011-2012 growing season. The developers measured levels of the introduced proteins by enzyme linked immunosorbent assay (ELISA) in leaves, roots, forage, and seed samples collected at multiple growth stages.
They did not sequence the produced protein as far as I can see, which would be the most direct answer to your question. But they did sequence the gene that was inserted, and they measured that the protein was produce by ELISA. This assay works by using antibodies that specifically recognize the target protein, together these data points strongly indicate that the desired protein is produced with the correct sequence.