LOAD vs OVERWRITE INSERT PRODUCE DIFFERENT TABLE SIZE!!

Question

Hi again!
Here is another issue I don't understand why happens:
The size of the table doubles if I load the data with INSERT OVERWRITE vs LOAD. As follows is an illustration of the problem:
I created a table "item". Loaded the data from item.dat (aprox 28MB). After that what happens is that the file item.dat will be moved to hive/warehouse and off course the size remains the same
Now if I create another table "item2" same as item and then load the data from item to item2 with the following command:
INSERT OVERWRITE TABLE item2 SELECT * FROM item
the size of table item2 is double of item (aprox 55MB)
Why does this happen? And is there any way to avoid it?
And the situation escalates as the size of the data grows.
ps. this is only to illustrate the problem. In practice I am interested for pre-joining tables but INSERT OVERWRITE increases the size of the joined table drastically (Actual problem: 4GB joined with 28MB gives 18GB)
Thank you!

Answer 1

Latest update on the issue:
I tested on cloudera as well and it is the same behavior: As follows are some details after running describe formatted <table_name> :
item (with LOAD):
33 Table Type: MANAGED_TABLE NULL
34 Table Parameters: NULL NULL
35 COLUMN_STATS_ACCURATE true
36 numFiles 1
37 numRows 0
38 rawDataSize 0
39 totalSize 28855325
40 transient_lastDdlTime 1427988576
41 NULL NULL
42 # Storage Information NULL NULL
43 SerDe Library: org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe NULL
44 InputFormat: org.apache.hadoop.mapred.TextInputFormat NULL
45 OutputFormat: org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat NULL
46 Compressed: No NULL
item2 (with INSERT OVERWRITE):
33 Table Type: MANAGED_TABLE NULL
34 Table Parameters: NULL NULL
35 COLUMN_STATS_ACCURATE true
36 numFiles 1
37 numRows 102000
38 rawDataSize 52058005
39 totalSize 52160005
40 transient_lastDdlTime 1427990208
41 NULL NULL
42 # Storage Information NULL NULL
43 SerDe Library: org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe NULL
44 InputFormat: org.apache.hadoop.mapred.TextInputFormat NULL
45 OutputFormat: org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat NULL
46 Compressed: No NULL
And now I don’t understand why the number of rows is 0 for item. When I query it returns 102000 rows