I have this schema
[ id | parent | inserted_on         | data    ],
[ 1  | 0      | 2017-05-04 09:29:37 | include ],
[ 2  | 1      | 2017-05-04 16:00:45 | include ],
[ 3  | 0      | 2017-05-04 16:01:13 | include ],
[ 4  | 1      | 2017-05-04 16:28:09 | include ],
[ 5  | 0      | 2017-05-04 16:28:11 | include ],
[ 6  | 1      | 2017-05-04 16:28:18 | include ],
[ 7  | 5      | 2017-05-04 16:28:23 | include ],
[ 8  | 1      | 2017-05-04 16:28:32 | exclude ],
[ 9  | 3      | 2017-05-04 16:28:39 | include ],
[ 10 | 0      | 2017-05-04 16:28:41 | include ]

And I need to make a virtual table which generate an `index` and `updated_on` column.
1. Select all data with parent = 0
2. Generate `updated_on` by this logic: `updated_on` = count children > 0 ? last children's inserted_on : inserted_on of itself; and then join it
3. Order the table by `updated_on` in descending order.
4. Index it

with this query:
SELECT @i := @i + 1 AS `index`, `p`.`id`, `p`.`parent`, `p`.`inserted_on`, `t`.`updated_on` FROM `datas` AS `p` JOIN (SELECT @i := 0) AS `iterator` LEFT JOIN (SELECT IF(`parent` = 0, `id`, `parent`) AS `timestamp_of`, MAX(`inserted_on`) AS `updated_on` FROM `datas` WHERE `data` != 'exclude' GROUP BY `timestamp_of`) AS `t` ON `p`.`id` = `t`.`timestamp_of` WHERE `p`.`parent` = '0' ORDER BY `t`.`updated_on` DESC;

I expect this result:
[ index | id | parent | timestamp           | data    | updated_on     DESC ],
[ 1     | 10 | 0      | 2017-05-04 16:28:41 | include | 2017-05-04 16:28:41 ],
[ 2     | 3  | 0      | 2017-05-04 16:01:13 | include | 2017-05-04 16:28:39 ],
[ 3     | 1  | 0      | 2017-05-04 09:29:37 | include | 2017-05-04 16:28:18 ],
[ 4     | 5  | 0      | 2017-05-04 16:28:11 | include | 2017-05-04 16:28:23 ]

but instead, I get:
[ index | id | parent | timestamp           | data    | updated_on     DESC ],
[ 4     | 10 | 0      | 2017-05-04 16:28:41 | include | 2017-05-04 16:28:41 ],
[ 2     | 3  | 0      | 2017-05-04 16:01:13 | include | 2017-05-04 16:28:39 ],
[ 3     | 1  | 0      | 2017-05-04 09:29:37 | include | 2017-05-04 16:28:18 ],
[ 1     | 5  | 0      | 2017-05-04 16:28:11 | include | 2017-05-04 16:28:23 ]

Where did I get it wrong?

Schema query:
CREATE TABLE `datas` (
  `id` int(11) NOT NULL,
  `parent` int(11) NOT NULL,
  `inserted_on` datetime NOT NULL,
  `data` varchar(16) COLLATE utf8_general_mysql500_ci NOT NULL
) ENGINE=MyISAM DEFAULT CHARSET=utf8 COLLATE=utf8_general_mysql500_ci ROW_FORMAT=DYNAMIC;
INSERT INTO `datas` (`id`, `parent`, `inserted_on`, `data`) VALUES
(1, 0, '2017-05-04 09:29:37', 'include'),
(2, 1, '2017-05-04 16:00:45', 'include'),
(3, 0, '2017-05-04 16:01:13', 'include'),
(4, 1, '2017-05-04 16:28:09', 'include'),
(5, 0, '2017-05-04 16:28:11', 'include'),
(6, 1, '2017-05-04 16:28:18', 'include'),
(7, 5, '2017-05-04 16:28:23', 'include'),
(8, 1, '2017-05-04 16:28:32', 'exclude'),
(9, 3, '2017-05-04 16:28:39', 'include'),
(10, 0, '2017-05-04 16:28:41', 'include');

[SOLVED]
The solution is to change LEFT JOIN to INNER JOIN.