Have you heard of the margin of error and subsampling?
So for example, if you poll people UK wide and ask them how they intend to vote you will get a picture that should be accurate within 3% plus or minus.
However, the margin of error will be much higher within certain regions because you will poll a much smaller number there. That is why the subsample of UK wide polls generally give a very poor picture of Scottish voting opinion because the margin of error is much higher within a subsample.
Explained in detail here:
anthonybmasters.medium.com/the-sub-sample-problem-ed15421c5d39
Scots make up 9% of the UK population, the LGBT population is roughly 5%. So the margin of error is even higher within that poll.
Furthermore, can you tell me whether it was even weighted to include LGBT people? Most polls are weighted by age, region, voting record and gender. Never seen one weighted with the LGBT population in mind. Can you confirm whether this is the case?