Weird result sets counts upon datetime criteria - sql

Question

I have the following table (the server is SQL Server 2016)
CREATE TABLE [dbo].[MyTable](
[RefID] [uniqueidentifier] NOT NULL,
[LG_ID] [bigint] NOT NULL,
[APP_SERVER_ID] [int] NOT NULL,
[WEB_SERVER_ID] [int] NOT NULL,
[WEB_BROWSER_ID] [int] NULL,
[CLIENT_TYPE_ID] [int] NOT NULL,
[OS_NAME_ID] [int] NULL,
[CLIENT_VERSION] [varchar](10) NULL,
[OS_VERSION] [varchar](10) NULL,
[BROWSER_VERSION] [varchar](10) NULL,
[DETAILS] [nvarchar](255) NULL,
[LG_SOURCE_TABLE] [char](2) NOT NULL,
[LG_TIME] [datetime] NULL,
CONSTRAINT [PK_MyTable] PRIMARY KEY CLUSTERED
(
[RefID] ASC,
[LG_ID] ASC,
[LG_SOURCE_TABLE] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [Primary]
) ON [Primary]
CREATE NONCLUSTERED INDEX [IX_MyTable_LG_TIME] ON [dbo].[MyTable]
(
[LG_TIME] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
GO
ALTER TABLE [dbo].[MyTable] ADD CONSTRAINT [PK_MyTable] PRIMARY KEY CLUSTERED
(
[RefID] ASC,
[LG_ID] ASC,
[LG_SOURCE_TABLE] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
GO
And I am querying data with the following methods (I am using only method 1, the other are tests that I am doing):
METHOD 1:
SELECT
[RefID],
[LG_ID],
[APP_SERVER_ID],
[WEB_SERVER_ID],
[WEB_BROWSER_ID],
[CLIENT_TYPE_ID],
[OS_NAME_ID],
[CLIENT_VERSION],
[OS_VERSION],
[BROWSER_VERSION],
[DETAILS],
[LG_SOURCE_TABLE],
[LG_TIME]
FROM
[MyTable] WITH (NOLOCK)
WHERE
[LG_TIME] >= '20190923'
AND [LG_TIME] < '20190924'
--returns 19.880.317 rows
METHOD 2:
SELECT
[RefID],
[LG_ID],
[APP_SERVER_ID],
[WEB_SERVER_ID],
[WEB_BROWSER_ID],
[CLIENT_TYPE_ID],
[OS_NAME_ID],
[CLIENT_VERSION],
[OS_VERSION],
[BROWSER_VERSION],
[DETAILS],
[LG_SOURCE_TABLE],
[LG_TIME]
FROM
[MyTable] WITH (NOLOCK)
WHERE
CAST([LG_TIME] AS DATE) = '20190923'
--returns 19.879.488 rows
METHOD 3:
SELECT
*
FROM
[MyTable] WITH (NOLOCK)
WHERE
[LG_TIME] >= '20190923'
AND [LG_TIME] < '20190924'
--returns 19.880.186 rows
METHOD 4:
SELECT
*
FROM
[MyTable] WITH (NOLOCK)
WHERE
CAST([LG_TIME] AS DATE) = '20190923'
--returns 19.879.488 rows
And as you can see I am having different result sets each time.
I don't understand why this is happening, am I doing something wrong?
Which is the correct/safe way to get the data I need based on datetime range?
I am doing a bcp out/in from production server (which has 1.2 billion rows but will be truncated soon) to an archive (which has 10.5 billion rows) and sometimes I am losing a record because they have the same [LG_TIME] (maybe I will create a different question about that).
EDIT 1: Some extra info:
The query (METHOD 1) is part of BCP OUT routine that happens daily at about 00:05 and the date range is always the same, 2 current days ago.
I have strict guidelines to always use WITH (NOLOCK) at queries at production servers.
There are millions of rows that have LG_TIME = NULL at the table but these rows are obsolete and will be deleted, i don't know if these rows cause any issues with my query.
The table is part of logging mechanism of the application, only INSERTS are happening and DELETEs very early morning (about 02:00-03:00) when the daily archiving is success. No one else is reading the table that time.
EDIT 2 (2019-09-26 10:30):
So i added LG_TIME IS NOT NULL to all my queries at where criteria and these are my results:
METHOD 1: 19.879.488 rows
METHOD 2: 19.879.488 rows
METHOD 3: 19.880.036 rows
METHOD 4: 19.879.488 rows
Still weird!