In ms recording, I think it goes like this:
ch. 1 = centre pointing card = (L+R)
ch. 2 = sideways figure of eight = (L-R)
Instead you ran cards pointed left and right. The ms process left you with:
ch.1 = (ch1 + ch2) = (L+R) which is additive and will be a hot signal
ch.2 = (ch1 - ch2) = (L- R) which is subtractive, sounds like crap and is very low level
You want left & right info back. I think that if you process thru a software ms decoder you will get this.
ms output left = (L+R)+(L-R) = 2L this is your left channel*
ms output right = (L+R)-(L-R) = 2R this is your right channel
The trick is when you mixed the original there was a certain balance encoded into the output (*it wasn't the pure channel as indicated above because they were mixed at different levels). You'll have to play with the width to try and find the right balancing spot. I think you can locate that spot by centering the sound. When you're off balance with the 'reverse ms decoding' one channel should get louder than the other (I think the left goes down as you add side channel).
All theoretical of course but try it out.
If you can't get it to sound ok then try reversing which channel you call mid and side.