On reading it I see that I have misrepresented the figures from memory:
2.88% of regional/list ballots were invalid
4.075% of constituency ballots were invalid
1.83% of local government ballots were invalid
These were against a historical spoilt paper rate of about 0.66%
(I have edited the original post to correct it)
The regional and constituency ballots were on the same physical piece of paper and if you voted a full ticket (eg SNP/SNP or Labour/Labour) there was only one way to do it.
For small parties (Greens, SSP) which only ran on the regional list you had to split the ticket. And there was one valid way to vote Labour/Green and one invalid way - so the small parties were much more liable to get invalid votes. The number of independent/small party MSP's was lower than expected.