This would be an area in which I have some academic expertise (hi, pilot!). The ETSI standards for GSM in Europe (to which network operators, service providers and handset manufacturers all work) specify far less rigorous protocols for selecting base stations for SMS or MMS transmission than for voice calls. Bolint had indeed posted some of the selection protocols, but those are the ones that apply to voice calls.
There are two main reasons why SMS and MMS can and should be dealt with differently: these communications are simplex (i.e. they are only one-way communications, from the base station to the handset, with no need for simultaneous transmission from the handset to the base station as would be the case with a normal voice call); and they are asynchronous (i.e. the information does not have to be delivered at a strictly-timed and synchronised basis - unlike voice calls which have to be transmitted and received in real time, in exactly the right order, and with minimal time delay).
Because of these two important differences, base station selection protocols are far less rigorous with SMS/MMS than with voice. When an SMS/MMS is sent or received, it's actually fairly common for the signal path to involve a more "distant" base station - leaving the base station with the strongest signal path free for more complex, duplex, synchronous voice calls. It's therefore far from unlikely for Knox's handset to have been situated in Sollecito's apartment, but to have received Lumumba's SMS from a base station which was not the optimal one for that location. The appeals documents allude to this, and I would hope that Hellmann would allow expert witnesses to appear before the court and explain all this properly.