Why is the pointcloud during mapping separated like grids, and there are very few points in each frame

[handshow888]

Figure 1 and 2 are the pcd I saved. It looks like many points are clustered into grids.
Figure3 is the screenshot of each frame from rviz2, I subscribe the /cloud_registered and set Size(pixels) to 5(default:1)， I only managed to see the points after that. No matter which angle I rotate the lidar to, the number of points is always small and sparse.
In figure4, I set Size(pixels) to 2, Decay Time to 300, it can be seen that although it is a smooth wall, there are still some points that are particularly dense, while others are very sparse.

Lidar: livox mid360
Due to special requirements, I flipped the roll axis of the radar 180 degrees.
I run ig_lio_mapping.launch.py, and it loaded livox.yaml.
here is my params:

/**:
  ros__parameters:
    #for odom
    odom/sendTF: true
    #topic names 
    odom/lidar_topic: /livox/lidar #/points_raw
    odom/imu_topic: /livox/imu #/imu_correct
    odom/lidar_type: livox # livox velodyne ouster

    #frames
    odom/odom_frame: lio_odom
    odom/robot_frame: base_link
    odom/imu_frame: imu_link
    odom/lidar_frame: livox_frame
    odom/min_radius: 1.0
    odom/max_radius: 150.0
    odom/point_filter_num: 1
    odom/time_scale: 1000.0 # nclt: 0.001 other: 1000.0
    odom/enable_ahrs_initalization: false
    odom/enable_acc_correct: true

    odom/scan_resolution: 0.05
    odom/voxel_map_resolution: 0.5
    odom/max_iterations: 10

    odom/acc_cov: 0.1
    odom/gyr_cov: 0.1
    odom/ba_cov: 0.000001   #0.000001
    odom/bg_cov: 0.000001   #0.000001
    odom/init_ori_cov: 0.0001
    odom/init_pos_cov: 0.0001
    odom/init_vel_cov: 100.0
    odom/init_ba_cov: 0.0001
    odom/init_bg_cov: 0.0001
    odom/gravity: 9.80665

    odom/gicp_constraints_gain: 100.0 #100.0
    odom/point2plane_constraints_gain: 1000.0 #1000.0
    odom/enable_undistort: true
    odom/enable_outlier_rejection: true

    

    
    #for extrinsics
    extrinsics/imu2lidar/t: [0.011, -0.02329, 0.04412]
    extrinsics/imu2lidar/r: [1.0, 0.0, 0.0,
                            0.0, 1.0, 0.0,
                            0.0,  0.0, 1.0 ]
    
    extrinsics/robot2lidar/t: [0.0, 0.0, 0.0]
    extrinsics/robot2lidar/r: [1.0, 0.0, 0.0,
                              0.0, 1.0, 0.0,
                              0.0,  0.0, 1.0 ]
                              
    extrinsics/robot2imu/t: [0.0, 0.0, 0.0]
    extrinsics/robot2imu/r: [1.0, 0.0, 0.0,
                            0.0, 1.0, 0.0,
                            0.0,  0.0, 1.0 ]

    #for map
    map/save_map_path: /home/handshow/ig_lio/map #do not put / at the end of line
    map/map_name: loc
    map/map_frame: map  # lio_odom

[handshow888]

by the way, I use the ros2 version from pr