LSTM using previously predicted data to attain new values converges

Question

I am trying to compare the output of an LSTM to my baseline data of an assets price t+100 into the future using a rolling frame, for example:

I take 5 consecutive price points, from date n1 to n5, I predict n6, and push n6 into the LSTM input data and remove n1 and repeat over until I have exausted my data making sure that each time I remove n1, I store this in a 'tracking' list to compare against my known data t+100 into the future. The idea is to keep predicting values using 5 previsouly predicted data points.

What I have noticed is that very quickly my predicted n6 values start to converge to a single value, almost straight away, see fig below:

My current configuration for the LSTM looks like so:

model = Sequential()
model.add(LSTM(100, activation='swish', input_shape=(5, 1), return_sequences=True))
model.add(LSTM(50, activation = 'swish'))
model.add(Dense(1))
model.compile(optimizer='adam', loss='mse', metrics = 'mse')
model.fit(X,y,epochs=100)

When I run the model against a rolling 5 point timeframe of the test data plotting the predicted values, I get a very accurate match to the test data set which suggests my model is okay.

I was expecting some arbitraty values to come out of predicting price points between t and t+100 but as for why it converges in figure 1 I am not entirely sure, is this an expected outcome of using predicted data over and over to predict more data?

Edit: Here is an example of payload data showing the shifting values with time:

[[[0.86381593]
  [0.81616699]
  [0.82052763]
  [0.81930565]
  [0.82811061]]]
round: 1
[[[0.81616699]
  [0.82052763]
  [0.81930565]
  [0.82811061]
  [0.81889451]]]
round: 2
[[[0.82052763]
  [0.81930565]
  [0.82811061]
  [0.81889451]
  [0.81397408]]]
round: 3
[[[0.81930565]
  [0.82811061]
  [0.81889451]
  [0.81397408]
  [0.80748832]]]
round: 4
[[[0.82811061]
  [0.81889451]
  [0.81397408]
  [0.80748832]
  [0.79987484]]]
round: 5
[[[0.81889451]
  [0.81397408]
  [0.80748832]
  [0.79987484]
  [0.79139411]]]
round: 6
[[[0.81397408]
  [0.80748832]
  [0.79987484]
  [0.79139411]
  [0.7828725 ]]]
round: 7
[[[0.80748832]
  [0.79987484]
  [0.79139411]
  [0.7828725 ]
  [0.77399129]]]
round: 8
[[[0.79987484]
  [0.79139411]
  [0.7828725 ]
  [0.77399129]
  [0.76494324]]]
round: 9
[[[0.79139411]
  [0.7828725 ]
  [0.77399129]
  [0.76494324]
  [0.75581664]]]
round: 10
[[[0.7828725 ]
  [0.77399129]
  [0.76494324]
  [0.75581664]
  [0.74665457]]]