The following transform should do what you want,
However if you have the following source data
| Col1 | Col2 | Col3 |
| 05/09/1999 | value 1 | 45345 |
| 01/06/2003 |
| 345 |
| Value 2 | 345 |
and you set Col1 up as a Date, then when it reads the empty value it tries to cast it to a Date - which fails.
Having reviewed this functionality we think it should return a 'null' value when no data is present in a cell, which would see the above transform generating the data
<Root>
<Row>
<MyDate>1999-09-05</MyDate>
<MyString>value 1</MyString>
<MyInt>45345</MyInt>
</Row>
<Row>
<MyDate>2003-06-01</MyDate>
<MyString></MyString>
<MyInt>345</MyInt>
</Row>
<Row>
<MyString>Value 2</MyString>
<MyInt>345</MyInt>
</Row>
</Root>
We will make the changes in the next point release.
In the meantime you can do this to work around the issue
Where Col1 is defined as containing a "string".