using LSD you test 2 means at 5% prob; 3 means at 12.2% Prob; 4 means at 20.3%prob.; 5 means at 28.6% prob etc (see attached file: I wrote this for my Scientists in 1978 Annual Report of Grassland Research Institute, Hurley, UK)
Why would you think that you can compare only four means with LSD?
Other than the inflation of type-I errors mentioned by Mewa Singh Dhanoa ?
By the way, this article talks about a modication to LSD that uses the Studentized range to better protect against type-I errors. https://www.utd.edu/~herve/abdi-LSD2010-pretty.pdf .
To maintain your testing probability at 5% (that is what that studentized Table does by enlarging the t-values according to how many means you are comparing; This is the agreed statistical proceedure)