I am sure I have included this before in a reply, however I think it is worth stating again.
The variation in MDD which is achieved using either the 2.5kg (standard) or 4.5kg (modified) will be dependant upon the soil type, however, as a rule of thumb, in the UK cohesive soils generally have a MDD suing the 2.5kg rammer close to 95% of the MDD for the 4.5 kg rammer. With alluvial silts and sands, the numbers can be very similar as there is a paucity of clay fines, and the materials in it self very spherical (like compacting marbles).
As to the difference between 90% and 95%, then the easiest way to get a handle on this is to draw out the graph of the moisture content/dry density relationship, then add the horizontal lines equivalent to 95% and 90% of the MDD. In addition to this, add in the air void lines. The range of materials which can meet the 90% requirement as opposed to the 95% is obviously much larger, has a significantly larger range in moistures and also much higher air void contents. Risk is as follows:
1. the increase in moisture permissible between 95% and 90% will reduce the bearing capacity/shear strength for cohesive soils and will increase the long-term settlement (self-weight, creep and consolidation) of the fill;
2. the increase in air void content permissible by increasing the range of densities from 95% to 90% will make the fill more susceptible to collapse and inundation settlement.
Therefore the difference is a 'potential' reduction in end-performance, reduction in bearing capacity, increase in settlement etc... In addition to this, if the material is cohesive, and is placed wetter than the 95% limit to the 90% limit, the trafficability of the fill will be greatly reduced and you will run the risk of the fill being unsuitable for placement. Again, based on experience with soils in the UK, fill placed above the moisture equivalent to 95% MDD will have a CBR of less than 2.5%, the lower permissible limit for engineered fill (in the UK according to the SHW and IAN 73 [HD 25])