Excel and R use the same RNG by default (Mersene-Twister), so you should get the same results for both programs. Excel's RAND() function uses the system time as a random seed, but I believe the toolpak uses the VBA function RND which can accept a seed for reproducibility.
If it's important to you to have the results match between the two, the first step in debugging would be to simplify. Set a seed and generate 3 random uniform numbers between 0 and 1 on both. They should match. Then, try 3 standard normal values, again, I would expect them to match.
If everything is good at that point, I would look at the code to see if you have the same structure, e.g. you're doing the same randomizations the same number of times, in the same order, etc.
After that, there may be more complicated issues at play.