Why there is different size of output coming from Sequence Folding Layer?

I wanted to design my own custom layer and I come across this problem.

Above are my Training data and the part of network with custom Layer which does nothing.

Below is the option I have set.

Though most of the outputs are as expected, there are some output as the below one as well from the flatten Layer.

Please let me know why this is caused?

Below I have also provided the script of Nothing Layer in the network

classdef nothingLayer < nnet.layer.Layer
    % Example custom PReLU layer.
    properties
       Alpha 
    end
   
    methods
        function layer = nothingLayer(name) 
            % layer = preluLayer(numChannels, name) creates a PReLU layer
            % with numChannels channels and specifies the layer name.

            % Set layer name.
            layer.Name = name;

            % Set layer description.
            layer.Description = "Split with " + " channels";
            
            % Initialize scaling coefficient.
            layer.Alpha = 1;
        
        end
        
        function Z = predict(layer, X)
            try
            validateattributes( X, { 'numeric' }, { 'vector', 'numel', 322, 'size',[1,322]});
            catch e
                display(e)
            end
            Z = layer.Alpha*X;
            display(Z);
        end
        
              
        function dLdX = backward(layer, X, ~, dLdZ, ~)
            %dLdX = zeros(size(X),'like',dLdZ);
            dLdX = layer.Alpha*dLdZ;
        end
    end