Hi Mansi,
I agree it is tempting to compare SQL and XML databases, but I would recommend you to refrain from doing so. My impression is that the systems are simply too different, so it will always be a comparison of apples and pears, oranges or even toothsticks.
If your data is regular enough to be stored in a simple table, if the data does not have any hierarchies, all values have fixed types, etc., you will never come close to a relational database, no matter which system you try. If you have hierarchies, things look different, but if you store such data relationally, it mainly gets interesting which data model you go for. The pure number of datasets or even nodes does not really matter that much, or it only begins to matter if all other factors have been clarified such that the results can be correctly interpreted.
Well, benchmarking is a huge field in itself. I clearly see that there is a requirement to compare both. I believe the practical approach is often more useful than a generic one:
* If you have real-life data, store this data in the systems of your choice * Formulate real-life queries that give you similar results * Compare these measurements to get an impression on the overall performance
Next, try to find bottlenecks. Optimize your data and queries for the specific systems. If you summarize the results, be careful about the presentation: "X is better than Y" sounds suspicious. Better say "given these datasets, these queries, these xyz, X seems to be...". I have once read an enthusiastic report about the performance of BaseX, claiming that it is much faster than X and Y, and obviously I first was pleased to read that, but as the objective of that report was to find one system that's better than all others, it was not helpful at all. But it is perfectly fine to say that system X proved, or feels, to be the better choice for a given scenario.
I hope my comments were not too lecturing.. Christian