The command sphrand creates a spherical uniform cloud, your method creates a square cloud. The distribution is different, random sure but the square cloud clumps results towards the corners of the square.
Explanation orientation has no magnitude so its point shaped. If you define it to have a constant size of 1 (normalized direction similar idea to surface normal), then the shape of the distribution is a sphere. Now the volume of the cloud defines the distributive property as seen form the center in this case. The distance form center to corner is longer than center to side, length in this case measures the relative likelihood. So its more likely that that random points to corner because there's more room for points there. (this does not mean they don't point elsewhere just more likely to point towards that direction)
This is why sphrand generates a more uniform random. When someone asks for random without specifying a distribution they usually mean uniform, mainly because its easier to make the data distributed form a uniform seed.
If this explanation does not suffice I'm happy to give a better explanation with visualization aids and proofs. Because in this case its very easy to do so.
