HMM Training for estimating transition and emission matrices

Hi all, I am trying to use the hmmtrain to learn the emission (emis) and transition (trans) matrices for a power dataset in dBm and subsequently generate data based on the emis and trans generated.

For now, I have discretized the noise power values, using integers to represent each bin of power values and to generate a sequence of emissions (seq) based on the discretized values. But after implementing the hmmtrain to generate an estimated trans and emis that depicts seq, I am unable to generate similar seqs using the estimated trans and emis (using hmmgenerate). In the sequence of states, hmmgenerate basically produces very few outcomes depicting state 1 leaving the rest in state 2. This does not correspond with my dataset which is about 800 and comprised of a whole lot more of state 1 outcomes than the model depicts. I need help/advice with this please.