Very good article. Too the point and easy to read.
A note to point out that although it's true that ROLAP is the slowest query performance of the different storage models, it's still possible to achieve second and subsecond response time with properly build aggregates.
I think it would have been good to disadvantages for each storage mode. For example although MOLAP is extremely fast it also cannot support the volume of data that ROLAP can. Plus ROLAP includes detail data, MOLAP will generall only store aggregates especially as the volume of data increases.
You also mention that with ROLAP users "will always be able to access the cube and retrieve real-time data." ROLAP is not cube technology, it's based on a relational database designed for reporting performance.
HOLAP sounds like it may be the best way to go... I'm not sure I have not been fortunate enough to work on one yet. Seems like you would get the best of both worlds, extremely fast aggregate data plus access to high volume detail data when it's needed.