What a strange requirement. Not sure what you mean by "old" and "new" identity values. Is the "old" value based on the value of col2? If not this seems completely pointless. The identity value here is what allows to uniquely define a row. Can you explain what the reason for this is? Especially since you are going to delete this table and then populate it with the last two inserted values I question why you even need it in a table. Why not just create a view instead?

select top 2 col1

,col2

from test1

order by col1 desc

This is not prone to error and will ALWAYS be accurate.