Hello Shivlu,
when having multiple ASBRs redistributing the same prefixes is better to use external type 1.
E2 metric is not added to the IGP metric to the ASBR.
E1 instead will compute the metric as the sum of the seed metric and the metric to ASBR router including intra and inter-area costs.
O E1 are preferred over O E2 routes even with a greater seed metric becuse they are more detailed.
So these strange results can derive from this.
I would suggest you to repeat your tests using O E1 routes for net 20.20.20.20/32.
Hope to help
Giuseppe